Nikolay I. Kolev Multiphase Flow Dynamics 1
Nikolay I. Kolev
Multiphase Flow Dynamics 1 Fundamentals
3rd Edition
With 149 Figures and CD-ROM
Nikolay Ivanov Kolev, PhD., DrSc. Möhrendorferstr. 7 91074 Herzogenaurach Germany
[email protected] Library of Congress Control Number: 2007923416
ISBN 978-3-540-69832-6 ISBN 978-3-540-22106-7
Springer Berlin Heidelberg New Y ork 2nd E dition Springer Berlin Heidelberg New Y ork
This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilm or in other ways, and storage in data banks. Duplication of this publication or parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and permission for use must always be obtained from Springer-Verlag. Violations are liable to prosecution under German Copyright Law. Springer is a part of Springer Science + Business Media springer.com
© Springer-Verlag Berlin Heidelberg 2002, 2005 and 2007 The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use.
Production: Integra Software Services Pvt. Ltd., Pondicherry, India Cover design: E rich Kirchner, Heidelberg Printed on acid-free paper
SPIN: 11926108
45/3100/Integra
543210
To Iva, Rali and Sonja with love!
The Balkan in Bulgaria, July. 2005, Nikolay Ivanov Kolev, 48x36cm oil on linen
A few words about the third extended edition
The reader will find in the third edition the information already included in the second one improved and extended in several places. Only the differences will be mentioned here. Chapter 1 contains brief remarks on the cinematic velocity of density wave propagation in porous structures and on diffusion term of the void propagation in case of pooling all the mechanical interactions in such kind of formalism. Chapter 2 contains collection of constitutive relations for the lift- and virtual mass forces, for the wall boundary layer forces, for the forces causing turbulent diffusion and for the forces forcing the rejection of a droplet deposition at wall with evaporation. Chapter 3 contains additional information on the construction of the saturation line by knowing pressure or temperature. An application of the material given in Chapter 3 is given in the new Volume 3 of this work to diesel fuel, where inherently consistent set of equation of state for both, gas and liquid is formulated. In addition a section was provided defining the equilibrium of the gases dissolved in liquids. These basics are then used in Volume 3 to construct approximation for the equilibrium solution concentrations of H2, O2, N2 and CO2 in water and to describe the non-equilibrium solution and dissolution at bubbles-, droplets- and films interfaces which extend the applicability of the methods of the multiphase fluid dynamics to flows with non-equilibrium solution and dissolution of gasses. An additional appendix to Chapter 3 a table with the partial derivatives of different forms of the equation of state is provided. Chapter 4 contains in addition a careful state of the art review for the application of the method of characteristics for modeling one and 2D-flows in the engineering practice. Chapter 5 contains an additional example for computation of the irreversible viscous dissipation in the boundary layer. For easy application additional Sections are added to Chapter 5 containing the different notations of the energy conservation for lumped parameters volumes and steady state flows. The limiting case with a gas flow in a pipe is considered in order to show the important difference to the existing gasdynamics solution in case that the irreversible heat dissipation due to friction is correctly taken into account. Due to the large number of the color pictures in Chapters 14 of this volume and Chapter 26 of Volume 2 they are included as a PDF-files in the attached CD. Chapter 26 of Volume 2 is completely rewritten. First, it contains the quick look of the IVA computer code with several interesting demonstration of the power of the technology described in this three Volumes work. A list of references is provided documenting the IVA-code development and validation. Review of the state of the art of the instability analyses of boiling systems is provided and interesting comparison of the modern IVA-predictions with large scale AREVA- experiments
VIII
A few words about the third extended edition
is provided. Powerful demonstration of the methods for analysing pressure wave propagations in single and two-phase systems is given by comparison with the 1983-Interatome experiments performed on simple and complex pipenetworks. Section 26.15 contains comparison with 333 experiments for variety of bundles, flow regimes including dry out, steady state and transients. It clearly demonstrates the power of the method and its well-defined uncerntainty to simulate boiling processes in complex geometry. And finally, Section 26.21 contains the discussion regarding: Is it possible to design universal multiphase flow analyzer? It contains my personal vision for future development of the multiphase fluid dynamics. Erlangen, September 2006
Nikolay Ivanov Kolev
A few words about the second extended edition
The reader will find in the second edition the information already included in the first one improved and extended in several places. Chapter 3 of Volume I has been completely rewritten. It now contains the next step in the generalization of the theory of the equations of states for arbitrary real mixtures. Now with one and the same formalism a mixture of miscible and immiscible components in arbitrary solid, liquid or gaseous states mixed and/or dissolved can be treated. This is a powerful method towards creating a universal flow analyzer. Chapter 6 has been extended with cases including details of modeling of combustion and detonation of hydrogen by taking into account the equilibrium dissociation. In Chapter 9, dealing with detonation during melt-water interaction, additional introductory information is given for the detonation of hydrogen in closed pipes taking into account the dissociation of the generated steam. A new Chapter 11 is inserted between the former Chapters 10 and 11 giving the mathematical tools for computing eigenvalues and eigenvectors and for determination of the type of systems of partial differential equations. The procedure for transformation of a hyperbolic system into canonical form is also provided. Then the relations between eigenvalues and critical flow and between eigenvalues and propagation velocity of small perturbation are briefly defined. This is in fact a translation of one chapter of my first book published in German by Springer in 1986. In Chapter 12 about the numerical solution methods, the variation of the volume-porosity with time is systematically incorporated into the numerical formalism. Appendix 2 of Volume I contains some additional information about orthogonal grid generation. Chapter 26 of Volume II, which is included in the accompanying CD, contains some additional experiments and movies documenting the performance of the method for fast pressure wave propagation in 2D geometry and interesting acoustical problems of melt-water interaction. Of course misprints and some layout deficiencies have also been removed as is usual for a second edition of such voluminous material. The form of the second improved and extended edition has been reached after I received many communications from all over the world from colleagues and friends commenting on different aspects of the two volumes or requesting additional information. I thank all of you who have contributed in this way to improving the two volumes. Erlangen, February 2004
Nikolay Ivanov Kolev
Summary
This monograph contains theory, methods and practical experience for describing complex transient multi-phase processes in arbitrary geometrical configurations. It is intended to help applied scientists and practicing engineers to understand better natural and industrial processes containing dynamic evolutions of complex multiphase flows. The book is also intended to be a useful source of information for students in the high semesters and in PhD programs. This monograph consists of three volumes: Vol. 1 Fundamentals (14 Chapters and 2 Appendixes), 750 pages + 322 paged + movies on CD-ROM Vol. 2 Mechanical and thermal interactions (26 Chapters), 912 pages Vol. 3 Selected chapters: turbulence, gas absorption and release by liquid, diesel fuel properties, 300 pages In Volume 1 the concept of three-fluid modeling is introduced. Each of the fields consists of multi-components grouped into an inert and a non-inert components group. Each field has its own velocity in space and its own temperature allowing mechanical and thermodynamic non-equilibrium among the fields. The idea of dynamic fragmentation and coalescence is introduced. Using the Slattery-Whitaker local spatial averaging theorem and the Leibnitz rule, the local volume-averaged mass, momentum and energy conservation equations are rigorously derived for heterogeneous porous structures. Successively a time averaging is performed. A discussion is provided on particle size spectra and averaging, cutting off the lower part of the spectrum due to mass transfer, the effect of the averaging on the effective velocity difference etc. In the derivation of the momentum equations special attention is paid to rearranging the pressure surface integrals in order to demonstrate the physical meaning of the originating source terms in the averaged systems and their link to hyperbolicity. The Reynolds stress concept is introduced for multi-phase flows. Before deriving the energy conservation in Chapter 5, I provide a Chapter 3 in which it is shown how to generate thermodynamic properties and the substantial derivatives for different kinds of mixtures by knowing the properties of the particular constituents. This chapter provides the necessary information to understand the entropy concept which is presented in Chapter 5. In the author's experience understanding the complex energy conservation for multi-phase systems and especially the entropy concept is very difficult for most students and practicing engineers. That is why Chapter 4 is provided as an introduction, showing the variety
XII
Summary
of notation of the energy conservation principle for single-phase multi-component flows. The local volume-averaged and time-averaged energy conservation equation is derived in Chapter 5 in different notation forms in terms of specific internal energy, specific enthalpy, specific entropy, and temperatures. The introduction of the entropy principle for such complex systems is given in detail in order to enable the practical use of the entropy concept. The useful “conservation of volume” equation is also derived. Examples for better understanding are given for the simple cases of lumped parameters – Chapter 6, infinite heat exchange without interfacial mass transfer, discharge of gas from a volume, injection of inert gas in a closed volume initially filled with inert gas, heat input in a gas in a closed volume, steam injection in a steam-air mixture, chemical reaction in a gas mixture in a closed volume and hydrogen combustion in an inert atmosphere. The exergy for a multiphase, multi-component system is introduced in Chapter 7 and discussed for the example of judging the efficiency of a heat pump. Simplification of the resulting system of PDEs to the case of one-dimensional flow is presented in Chapter 8. Some interesting aspects of the fluid structure coupling such as pipe deformation due to temporal pressure change in the flow, and forces acting on the internal pipe walls are discussed. The idea of algebraic slip is presented. From the system thus obtained the next step of simplification leads to the system of ordinary differential equations describing the critical multi-phase, multi-component flow by means of three velocity fields. Modeling of valves and pumps is discussed in the context of modeling of networks consisting of pipes, valves, pumps and other different components. Another case of simplification of the theory of multiphase flows is presented in Chapter 9, where the theory of continuum sound waves and discontinuous shock waves for melt-water interaction is presented. In order to easily understand it, the corresponding theory for single- and two-phase flows is reviewed as an introduction. Finally an interesting application for the interaction of molten uranium and aluminum oxides with water, as well of the interaction of molten iron with water is presented. Chapter 10 is devoted to the derivation of the conservation equations for multi-phase multi-component multi-velocity field flow in general curvilinear coordinate systems. For a better understanding of the mathematical basics used in this chapter two appendixes are provided: Appendix 1 in which a brief introduction to vector analysis is given and Appendix 2 in which the basics of the coordinate transformation theory are summarized. Chapter 11 describes numerical solution methods for different multi-phase flow problems. The first order donor-cell method is presented in detail by discretizing the governing equations, creating a strong interfacial velocity coupling, and strong pressure-velocity coupling. Different approximations for the pressure equations are derived and three different solution methods are discussed in detail. One of them is based on the Newton iterations for minimizing the residuals by using the conjugate gradients. A method for temperature inversion is presented. Several details are given enabling scientists and engineers to use this chapter for their own computer code development, such us integration procedure (implicit method), time step and accuracy control. Finally some highorder discretization schemes for convection-diffusion terms such us space exponential scheme and other high-order up-winding schemas are presented. Different analytical derivations are provided in Appendixes 11.1 to 11.8 including
Summary
XIII
the analytical derivatives of the residual error of each equation with respect to the dependent variables. Some important basic definitions that are required for describing a pipe networks are introduced. Chapter 12 presents a numerical solution method for multi-phase flow problems in multiple blocks of curvilinear coordinate systems generalizing in fact the experience gained by Chapter 11. Several important details of how to derive explicit pressure equations are provided. The advantage of using orthogonal grids also is easily derived from this chapter. Chapter 13 provides the mathematical tools for determination of the type of a system of partial differential equations. The procedures for computing eigenvalues and eigenvectors and the algorithm for transferring a hyperbolic system into canonical form are given. The relations between eigenvalues and critical flow and eigenvalues and propagation velocity of small perturbations are also defined. This completes the basics of the multi-phase, multi-component flow dynamics. Chapter 14 provides several numerical simulations as illustrations of the power of the methods presented in this monograph. A compact disc is attached that contains movies corresponding to particular cases discussed in this chapter. The movies can be played with any tool capable of accepting avi- or animated gif-files. Volume 2 is devoted to the so-called closure laws: the important constitutive relations for mechanical and thermal interactions. The structure of the volume has the character of a state-of-the-art review and a selection of the best available approaches for describing interfacial processes. In many cases the original contribution of the author is incorporated into the overall presentation. The most important aspects of the presentation are that it stems from the author’s long years of experience of developing computer codes. The emphasis is on the practical use of these relationships: either as stand alone estimation methods or within a framework of computer codes. Volume 3 is devoted to selected Chapters of the multiphase fluid dynamics that are important for practical applications: The state of the art of the turbulence modeling in multiphase flows is presented. As introduction, some basics of the single-phase boundary layer theory including some important scales and flow oscillation characteristics in pipes and rod bundles are presented. Then the scales characterizing the dispersed flow systems are presented. The description of the turbulence is provided at different level of complexity: simple algebraic models for eddy viscosity, algebraic models based on the Boussinesq hypothesis, modification of the boundary layer share due to modification of the bulk turbulence, modification of the boundary layer share due to nucleate boiling. Then the role of the following forces on the matematical description of turbulent flows is discussed: the lift force, the lubrication force in the wall boundary layer, and the dispersion force. A pragmatic generalization of the k-eps models for continuous velocity field is proposed containing flows in large volumes and flows in porous structures. Method of haw to derive source and sinks terms for multiphase k-eps models is presented. A set of 13 single- and two phase benchmarks for verification of k-eps models in system computer codes are provided and reproduced with the IVA computer code as an example of the application of the theory. This methodology is intendet to help other engineers and
XIV
Summary
scientists to introduce this technology step-by-step in their own engineering practice. In many practical application gases are solved in liquids under given conditions, released under other conditions and therefore affecting technical processes for good of for bad. There is almost no systematical description of this subject in the literature. That is why I decided to collect useful information on the solubility of oxygen, nitrogen, hydrogen and carbon dioxide in water under large interval of pressures and temperatures, provide appropriate mathematical approximation functions and validate them. In addition methods for computation of the diffusion coefficients are described. With this information solution and dissolution dynamics in multiphase fluid flows can be analyzed. For this purpose the nonequilibrium absorption and release on bubble, droplet and film surfaces under different conditions is mathematically described. In order to allow the application of the theory from the three Volumes also to processes in combustion engines a systematic set of internally consistent state equations for diesel fuel gas and liquid valid in broad range of changing pressure and temperature are provided.
Table of c ontents
Chapter 14 of volume 1 and Chapter 26 of volume 2 are available in pdf format on the CD-ROM attached to volume 1. The system requirements are Windows 98 and higher. Both pdf files contain links to computer animations. To see the animations, one double clicks on the active links contained inside the pdf documents. The animations are then displayed in an internet browser, such Microsoft Internet Explorer or Netscape. Alternatively, gif-file animations are also provided. Nomenclature.................................................................................................. XXV Introduction ................................................................................................. XXXV 1
Mass conservation ...........................................................................................1 1.1 Introduction .............................................................................................1 1.2 Basic definitions......................................................................................2 1.3 Non-structured and structured fields .......................................................9 1.4 Slattery and Whitaker’ s local spatial averaging theorem ......................10 1.5 General transport equation (Leibnitz rule) ............................................12 1.6 Local volume-averaged mass conservation equation ............................13 1.7 Time average.........................................................................................16 1.8 Local volume-averaged component conservation equations .................18 1.9 Local volume- and time-averaged conservation equations ...................20 1.10 Conservation equations for the number density of particles .................24 1.11 Implication of the assumption of mono-dispersity in a cell ..................30 1.11.1 Particle size spectrum and averaging........................................30 1.11.2 Cutting of the lower part of the spectrum due to mass transfer......................................................................................31 1.11.3 The effect of the averaging on the effective velocity difference..................................................................................33 1.12 Stratified structure .................................................................................35 1.13 Final remarks and conclusions ..............................................................35 References .......................................................................................................37
2 Momentums conservation.............................................................................41 2.1 Introduction ...........................................................................................41 2.2 Local volume-averaged momentum equations......................................42 2.2.1 Single-phase momentum equations ..........................................42 2.2.2 Interface force balance (momentum jump condition)...............42
XVI
Table of contents
2.2.3
Local volume averaging of the single-phase momentum equation.................................................................................... 49 2.3 Rearrangement of the surface integrals ................................................. 51 2.4 Local volume average and time average ............................................... 55 2.5 Dispersed phase in laminar continuum - pseudo turbulence ................. 57 2.6 Viscous and Reynolds stresses .............................................................. 57 2.7 Non-equal bulk and boundary layer pressures ...................................... 62 2.7.1 Continuous interface ................................................................ 62 2.7.2 Dispersed interface................................................................... 77 2.8 Working form for dispersed and continuous phase............................... 93 2.9 General working form for dispersed and continuous phases................. 98 2.10 Some practical simplifications ............................................................ 100 2.11 Conclusion .......................................................................................... 104 Appendix 2.1 ................................................................................................. 104 Appendix 2.2 ................................................................................................. 105 Appendix 2.3 ................................................................................................. 106 References ..................................................................................................... 109 3 Derivatives for the equations of state......................................................... 115 3.1 Introduction......................................................................................... 115 3.2 Multi-component mixtures of miscible and non-miscible components ......................................................................................... 117 3.2.1 Computation of partial pressures for known mass concentrations, system pressure and temperature................... 118 3.2.2 Partial derivatives of the equation of state ρ = ρ p, T , C2,...,imax .............................................................. 125
(
3.3
3.4
)
3.2.3
Partial derivatives in the equation of state T = T ϕ , p, C2,...,imax , where ϕ = s, h, e ................................. 130
3.2.4 3.2.5
Chemical potential.................................................................. 139 Partial derivatives in the equation of state ρ = ρ p, ϕ , C2,...,imax , where ϕ = s, h, e ................................. 150
(
(
)
)
Mixture of liquid and microscopic solid particles of different chemical substances ............................................................................ 153 3.3.1 Partial derivatives in the equation of state ρ = ρ p, T , C2,...,imax .............................................................. 153 3.3.2 Partial derivatives in the equation of state T = T p, ϕ , C2,...,imax where ϕ = h, e, s .................................. 154
(
)
(
)
Single-component equilibrium fluid ................................................... 155 3.4.1 Superheated vapor .................................................................. 155 3.4.2 Reconstruction of equation of state by using a limited amount of data available ........................................................ 156 3.4.3 Vapor-liquid mixture in thermodynamic equilibrium ............ 163
Table of contents
XVII
3.4.4 Liquid-solid mixture in thermodynamic equilibrium .............164 3.4.5 Solid phase .............................................................................164 3.5 Extension state of liquids ....................................................................165 Appendix 3.1 Application of the theory to steam-air mixtures......................165 Appendix 3.2 Useful references for computing properties of single constituents...........................................................................167 Appendix 3.3 Useful definitions and relations between thermodynamic quantities .............................................................................................169 References .....................................................................................................170 4 On the variety of notations of the energy conservation for single-phase flow....................................................................................173 4.1 Introduction .........................................................................................173 4.2 Mass and momentum conservation, energy conservation ...................174 4.3 Simple notation of the energy conservation equation .........................175 4.4 The entropy .........................................................................................176 4.5 Equation of state..................................................................................177 4.6 Variety of notation of the energy conservation principle ....................177 4.6.1 Temperature ...........................................................................177 4.6.2 Specific enthalpy ....................................................................178 4.7 Summary of different notations...........................................................179 4.8 The equivalence of the canonical forms..............................................179 4.9 Equivalence of the analytical solutions ...............................................182 4.10 Equivalence of the numerical solutions?.............................................183 4.10.1 Explicit first order method of characteristics..........................183 4.10.2 The perfect gas shock tube: benchmark for numerical methods ..................................................................................187 4.11 Interpenetrating fluids .........................................................................196 4.12 Summary of different notations for interpenetrating fluids.................201 Appendix 4.1 Analytical solution of the shock tube problem .......................203 Appendix 4.2 Achievable accuracy of the donor-cell method for single-phase flows .........................................................................207 References .....................................................................................................210 5 First and second laws of the thermodynamics ..........................................213 5.1 Introduction .........................................................................................213 5.2 Instantaneous local volume average energy equations........................216 5.3 Dalton and Fick’s laws, center of mass mixture velocity, caloric mixture properties ...............................................................................223 5.4 Enthalpy equation................................................................................225 5.5 Internal energy equation......................................................................229 5.6 Entropy equation .................................................................................230 5.7 Local volume- and time-averaged entropy equation ...........................234 5.8 Local volume- and time-averaged internal energy equation ...............239
XVIII
Table of contents
5.9 Local volume- and time-averaged specific enthalpy equation ............ 241 5.10 Non-conservative and semi-conservative forms of the entropy equation............................................................................................... 243 5.11 Comments on the source terms in the mixture entropy equation ........ 245 5.12 Viscous dissipation ............................................................................. 250 5.13 Temperature equation.......................................................................... 255 5.14 Second law of the thermodynamics .................................................... 259 5.15 Mixture volume conservation equation............................................... 260 5.16 Linearized form of the source term for the temperature equation ....... 265 5.17 Interface conditions............................................................................. 272 5.18 Lumped parameter volumes ................................................................ 273 5.19 Steady state ......................................................................................... 274 5.20 Final remarks....................................................................................... 278 References ..................................................................................................... 279 6 Some simple applications of the mass and energy conservation.............. 283 6.1 Infinite heat exchange without interfacial mass transfer ..................... 283 6.2 Discharge of gas from a volume ......................................................... 285 6.3 Injection of inert gas in a closed volume initially filled with inert gas......287 6.4 Heat input in a gas in a closed volume................................................ 288 6.5 Steam injection in a steam-air mixture................................................ 289 6.6 Chemical reaction in a gas mixture in a closed volume ...................... 292 6.7 Hydrogen combustion in an inert atmosphere..................................... 294 6.7.1 Simple introduction to combustion kinetics ........................... 294 6.7.2 Ignition temperature and ignition concentration limits .......... 296 6.7.3 Detonability concentration limits ........................................... 297 6.7.4 The heat release due to combustion ....................................... 297 6.7.5 Equilibrium dissociation ........................................................ 298 6.7.6 Source terms of the energy conservation of the gas phase ..... 303 6.7.7 Temperature and pressure changes in a closed control volume; adiabatic temperature of the burned gases................ 305 References ..................................................................................................... 309 7
Exergy of multi-phase multi-component systems .....................................311 7.1 Introduction......................................................................................... 311 7.2 The pseudo-exergy equation for single-fluid systems......................... 311 7.3 The fundamental exergy equation ....................................................... 313 7.3.1 The exergy definition in accordance with Reynolds and Perkins ............................................................................. 313 7.3.2 The exergy definition in accordance with Gouy (l’énergie utilisable, 1889) ..................................................... 314 7.3.3 The exergy definition appropriate for estimation of the volume change work .................................................... 315 7.3.4 The exergy definition appropriate for estimation of the technical work .............................................................. 316 7.4 Some interesting consequences of the fundamental exergy equation ....316
Table of contents
XIX
7.5
Judging the efficiency of a heat pump as an example of application of the exergy.................................................................318 7.6 Three-fluid multi-component systems.................................................320 7.7 Practical relevance...............................................................................323 References .....................................................................................................323 8 One-dimensional three-fluid flows .............................................................325 8.1 Summary of the local volume- and time-averaged conservation equations .............................................................................................325 8.2 Treatment of the field pressure gradient forces ...................................328 8.2.1 Dispersed flows ......................................................................328 8.2.2 Stratified flow.........................................................................329 8.3 Pipe deformation due to temporal pressure change in the flow...........329 8.4 Some simple cases...............................................................................331 8.5 Slip model – transient flow .................................................................338 8.6 Slip model – steady state. Critical mass flow rate...............................342 8.7 Forces acting on the pipes due to the flow – theoretical basics...........350 8.8 Relief valves........................................................................................356 8.8.1 Introduction ............................................................................356 8.8.2 Valve characteristics, model formulation ...............................357 8.8.3 Analytical solution .................................................................361 8.8.4 Fitting the piecewise solution on two known position – time points .............................................................363 8.8.5 Fitting the piecewise solution on known velocity and position for a given time ..................................................365 8.8.6 Idealized valve characteristics ................................................366 8.8.7 Recommendations for the application of the model in system computer codes.......................................................368 8.8.8 Some illustrations of the valve performance model ...............370 8.8.9 Nomenclature for Section 8.8.................................................376 8.9 Pump model ........................................................................................378 8.9.1 Variables defining the pump behavior....................................378 8.9.2 Theoretical basics ...................................................................381 8.9.3 Suter diagram .........................................................................388 8.9.4 Computational procedure .......................................................394 8.9.5 Centrifugal pump drive model................................................395 8.9.6 Extension of the theory to multi-phase flow...........................396 Appendix 1: Chronological references to the subject critical two-phase flow .....399 References .....................................................................................................405 9 Detonation waves caused by chemical reactions or by melt-coolant interactions...................................................................................................407 9.1 Introduction .........................................................................................407 9.2 Single-phase theory .............................................................................409 9.2.1 Continuum sound waves (Laplace) .......................................409 9.2.2 Discontinuum shock waves (Rankine-Hugoniot) ..................410
XX
Table of contents
9.2.3
The Landau and Liftshitz analytical solution for detonation in perfect gases............................................... 414 9.2.4 Numerical solution for detonation in closed pipes ................ 418 9.3 Multi-phase flow ................................................................................. 421 9.3.1 Continuum sound waves ....................................................... 421 9.3.2 Discontinuum shock waves ................................................... 423 9.3.3 Melt-coolant interaction detonations..................................... 424 9.3.4 Similarity to and differences from the Yuen and Theofanous formalism........................................................... 429 9.3.5 Numerical solution method ................................................... 430 9.4 Detonation waves in water mixed with different molten materials..... 431 9.4.1 UO2 water system.................................................................. 431 9.4.2 Efficiencies............................................................................ 435 9.4.3 The maximum coolant entrainment ratio............................... 438 9.5 Conclusions......................................................................................... 439 9.6 Practical significance .......................................................................... 441 Appendix 9.1 Specific heat capacity at constant pressure for urania and alumina......................................................................................... 442 References ..................................................................................................... 443 10 Conservation equations in general curvilinear coordinate systems ........ 445 10.1 Introduction......................................................................................... 445 10.2 Field mass conservation equations...................................................... 446 10.3 Mass conservation equations for components inside the field – conservative form................................................................................ 449 10.4 Field mass conservation equations for components inside the field – non-conservative form ............................................................. 451 10.5 Particles number conservation equations for each velocity field ........ 451 10.6 Field entropy conservation equations – conservative form ................. 452 10.7 Field entropy conservation equations – non-conservative form.......... 453 10.8 Irreversible power dissipation caused by the viscous forces............... 454 10.9 The non-conservative entropy equation in terms of temperature and pressure ........................................................................................ 456 10.10 The volume conservation equation ..................................................... 458 10.11 The momentum equations ................................................................... 459 10.12 The flux concept, conservative and semi-conservative forms............. 466 10.12.1 Mass conservation equation .................................................. 466 10.12.2 Entropy equation ................................................................... 468 10.12.3 Temperature equation............................................................ 468 10.12.4 Momentum conservation in the x-direction........................... 469 10.12.5 Momentum conservation in the y-direction........................... 470 10.12.6 Momentum conservation in the z-direction ........................... 472 10.13 Concluding remarks ............................................................................ 473 References ..................................................................................................... 473
Table of contents
XXI
11 Type of the system of PDEs ........................................................................475 11.1 Eigenvalues, eigenvectors, canonical form .........................................475 11.2 Physical interpretation.........................................................................478 11.2.1 Eigenvalues and propagation velocity of perturbations.........478 11.2.2 Eigenvalues and propagation velocity of harmonic oscillations.............................................................................478 11.2.3 Eigenvalues and critical flow ................................................479 References ...........................................................................................480 12 Numerical solution methods for multi-phase flow problems ...................481 12.1 Introduction .........................................................................................481 12.2 Formulation of the mathematical problem ..........................................481 12.3 Space discretization and location of the discrete variables .................483 12.4 Discretization of the mass conservation equations..............................488 12.5 First order donor-cell finite difference approximations.......................490 12.6 Discretization of the concentration equations .....................................492 12.7 Discretization of the entropy equation ................................................493 12.8 Discretization of the temperature equation..........................................494 12.9 Physical significance of the necessary convergence condition ...........497 12.10 Implicit discretization of momentum equations ..................................499 12.11 Pressure equations for IVA2 and IVA3 computer codes.....................505 12.12 A Newton-type iteration method for multi-phase flows ......................508 12.13 Integration procedure: implicit method ...............................................517 12.14 Time step and accuracy control...........................................................519 12.15 High order discretization schemes for convection-diffusion terms .....520 12.15.1 Space exponential scheme .....................................................520 12.15.2 High order upwinding ............................................................523 12.15.3 Constrained interpolation profile (CIP) method ....................525 12.16 Problem solution examples to the basics of the CIP method...............530 12.16.1 Discretization concept ...........................................................530 12.16.2 Second order constrained interpolation profiles ....................531 12.16.3 Third order constrained interpolation profiles .......................533 12.16.4 Fourth order constrained interpolation profiles .....................534 12.17 Pipe networks: some basic definitions ................................................554 12.17.1 Pipes ......................................................................................554 12.17.2 Axis in the space....................................................................556 12.17.3 Diameters of pipe sections.....................................................557 12.17.4 Reductions.............................................................................557 12.17.5 Elbows...................................................................................558 12.17.6 Creating a library of pipes .....................................................559 12.17.7 Sub system network...............................................................559 12.17.8 Discretization of pipes...........................................................560 12.17.9 Knots .....................................................................................560 Appendix 12.1 Definitions applicable to discretization of the mass conservation equations ........................................................................562 Appendix 12.2 Discretization of the concentration equations .......................565
XXII
Table of contents
Appendix 12.3 Harmonic averaged diffusion coefficients ............................ 567 Appendix 12.4. Discretized radial momentum equation ............................... 568 Appendix 12.5 The a coefficients for Eq. (12.46)....................................... 573 Appendix 12.6 Discretization of the angular momentum equation ............... 573 Appendix 12.7 Discretization of the axial momentum equation ................... 575 Appendix 12.8 Analytical derivatives for the residual error of each equation with respect to the dependent variables ................... 577 Appendix 12.9 Simple introduction to iterative methods for solution of algebraic systems ............................................................................ 580 References ..................................................................................................... 581 13 Numerical methods for multi-phase flow in curvilinear coordinate systems.......................................................................................................... 587 13.1 Introduction......................................................................................... 587 13.2 Nodes, grids, meshes, topology - some basic definitions.................... 589 13.3 Formulation of the mathematical problem .......................................... 590 13.4 Discretization of the mass conservation equations.............................. 592 13.4.1 Integration over a finite time step and finite control volume ...................................................................... 592 13.4.2 The donor-cell concept.......................................................... 594 13.4.3 Two methods for computing the finite difference approximations of the contravariant vectors at the cell center..................................................................... 597 13.4.4 Discretization of the diffusion terms ..................................... 599 13.5 Discretization of the entropy equation ................................................ 603 13.6 Discretization of the temperature equation ......................................... 604 13.7 Discretization of the particle number density equation ....................... 605 13.8 Discretization of the x momentum equation........................................ 605 13.9 Discretization of the y momentum equation........................................ 607 13.10 Discretization of the z momentum equation........................................ 608 13.11 Pressure-velocity coupling .................................................................. 608 13.12 Staggered x momentum equation ........................................................ 613 Appendix 13.1 Harmonic averaged diffusion coefficients ............................ 623 Appendix 13.2 Off-diagonal viscous diffusion terms of the x momentum equation............................................................................................... 625 Appendix 13.3 Off-diagonal viscous diffusion terms of the y momentum equation............................................................................................... 628 Appendix 13.4 Off-diagonal viscous diffusion terms of the z momentum equation............................................................................................... 630 References ..................................................................................................... 633 Appendix 1 Brief introduction to vector analysis ........................................... 637 Appendix 2 Basics of the coordinate transformation theory ......................... 663
Table of contents
XXIII
14 Visual demonstration of the method ..........................................................715 14.1 Melt-water interactions .......................................................................715 14.1.1 Cases 1 to 4 ...........................................................................715 14.1.2 Cases 5, 6 and 7.....................................................................721 14.1.3 Cases 8 to 10..........................................................................725 14.1.4 Cases 11 and 12.....................................................................736 14.1.5 Case 13 ..................................................................................739 14.1.6 Case 14 ..................................................................................740 14.2 Pipe networks ......................................................................................743 14.2.1 Case 15 ..................................................................................743 14.3 3D steam-water interaction .................................................................745 14.3.1 Case 16 ..................................................................................745 14.4 Three dimensional steam-water interaction in presence of non-condensable gases....................................................................746 14.4.1 Case 17 ..................................................................................746 14.5 Three dimensional steam production in boiling water reactor ............748 14.5.1 Case 18 ..................................................................................748 References .....................................................................................................749
Index ...................................................................................................................751
Nomenclature
Latin A A a
cross section, m² surface vector speed of sound, m / s
alw
surface of the field l wetting the wall w per unit flow volume
alσ
longing to control volume Vol (local volume interface area density of the structure w), m −1 surface of the velocity field l contacting the neighboring fields per unit
lmax
∑Vol l =1
flow volume
lmax
∑Vol l =1
l
l
be-
belonging to control volume Vol (local volume in-
terface area density of the velocity field l), m −1 al
Cui Cil c Cm Ci
cp vm
total surface of the velocity field l per unit flow volume
lmax
∑Vol l =1
l
belong-
ing to control volume Vol (local volume interface area density of the velocity field l), m −1 Courant criterion corresponding to each eigenvalue, dimensionless mass concentration of the inert component i in the velocity field l coefficients, dimensionless mass concentration of the component m in the velocity field, dimensionless mass concentration of the component i in the velocity field, dimensionless specific heat at constant pressure, J / ( kgK )
c cd cL Dhy
virtual mass force coefficient, dimensionless drag force coefficient, dimensionless lift force coefficient, dimensionless hydraulic diameter (4 times cross-sectional area / perimeter), m
D3E
diameter of the entrained droplets, m size of the bubbles produced after one nucleation cycle on the solid structure, bubble departure diameter, m
Dld
XXVI
Nomenclature
D1dm
Dill
size of bubbles produced after one nucleation cycle on the inert solid particles of field m = 2, 3 critical size for homogeneous nucleation, m critical size in presence of dissolved gases, m most probable particle size, m characteristic length of the velocity field l, particle size in case of fragmented field, m coefficient of molecular diffusion for species i into the field l, m 2 / s
Dilt
coefficient of turbulent diffusion, m 2 / s
Dil*
total diffusion coefficient, m 2 / s right-hand side of the non-conservative conservation equation for the
Dlch Dlcd Dl′ Dl
DCil
(
inert component, kg / sm 3 D d E e F (ξ )
)
2
diffusivity, m / s total differential total energy, J specific internal energy, J/kg function introduced first in Eq. (42) Chapter 2
F , f (... function of (...
f f Flw
force per unit flow volume, N / m3 fraction of entrained melt or water in the detonation theory surfaces separating the velocity field l from the neighboring structure
Flσ
within Vol, m 2 surfaces separating the velocity field l from the neighboring velocity field
F f im
within Vol, m 2 surface defining the control volume Vol, m 2 frequency of the nuclei generated from one activated seed on the particle
flw
belonging to the donor velocity field m, s −1 frequency of the bubble generation from one activated seed on the chan-
fl ,coal
nel wall, s −1 coalescence frequency, s −1
g H h hi I i J j
acceleration due to gravity, m / s 2 height, m specific enthalpy, J/kg eigenvectors corresponding to each eigenvalue unit matrix, dimensionless unit vector along the x-axis matrix, Jacobian unit vector along the y-axis
Nomenclature
XXVII
k k k kilT
unit vector along the k-axis cell number kinetic energy of turbulent pulsation, m2 / s 2 coefficient of thermo-diffusion, dimensionless
kilp L Mi m n ΔV n n le
coefficient of baro-diffusion, dimensionless length, m kg-mole mass of the species i, kg/mole total mass, kg unit vector pointing along ΔVml , dimensionless unit vector pointing outwards from the control volume Vol, dimensionless unit surface vector pointing outwards from the control volume Vol unit interface vector pointing outwards from the velocity field l
n lσ nl
number of the particle from species i per unit flow volume, m −3 number of particles of field i per unit flow volume, particle number den-
ncoal
sity of the velocity field l, m −3 number of particles disappearing due to coalescence per unit time and
nl , kin
unit volume, m−3 particle production rate due to nucleation during evaporation or conden-
nil
(
sation, 1/ m3 s nlw′′ nlh
) −2
number of the activated seeds on unit area of the wall, m number of the nuclei generated by homogeneous nucleation in the donor
(
velocity field per unit time and unit volume of the flow, 1/ m3 s nl , dis
(
field per unit time and unit volume of the flow, 1/ m3 s nl , sp
)
number of the nuclei generated from dissolved gases in the donor velocity
)
number of particles of the velocity field l arising due to hydrodynamic
(
disintegration per unit time and unit volume of the flow, 1/ m3 s P P Per pli p q ′′′ qσ′′′l
)
probability irreversibly dissipated power from the viscous forces due to deformation of the local volume and time average velocities in the space, W / kg perimeter, m l = 1: partial pressure inside the velocity field l l = 2,3: pressure of the velocity field l pressure, Pa thermal power per unit flow volume introduced into the fluid, W / m3 l = 1,2,3. Thermal power per unit flow volume introduced from the interface into the velocity field l, W / m3
XXVIII
qw′′′σ l
Nomenclature
thermal power per unit flow volume introduced from the structure inter-
face into the velocity field l, W / m3 R mean radius of the interface curvature, m r(x,y,z) position vector, m R (with indexes) gas constant, J/(kgK) s arc length vector, m S total entropy, J/K s specific entropy, J/(kgK) t Sc turbulent Schmidt number, dimensionless Sc tn is the turbulent Schmidt number for particle diffusion, dimensionless T temperature, K Tl temperature of the velocity field l, K shear stress tensor, N / m 2 unit tangent vector dependent variables vector control volume, m3 size of the control volume, m volume available for the field l inside the control volume, m3
T t U Vol Vol1/ 3 Voll lmax
∑Vol
volume available for the flow inside the control volume, m3
V
δ iVlτ
instantaneous fluid velocity with components, u, v, w in r , θ , and z direction, m/s instantaneous field velocity with components, ulτ , vlτ , wlτ in r ,θ , and z direction, m/s time-averaged velocity, m/s pulsation component of the instantaneous velocity field, m/s Vl − Vm , velocity difference, disperse phase l, continuous phase m carrying l, m / s diffusion velocity, m/s
Vlτσ
interface velocity vector, m/s
l =1
Vlτ Vl Vl′ ΔVlm
τ
Vl γ
v
x y ×
l
instantaneous vector with components, ulτ γ r , vlτ γ θ , wlτ γ z in r ,θ , and z directions, m/s specific volume, m3 / kg mass fraction, dimensionless distance between the bottom of the pipe and the center of mass of the liquid, m vector product
Nomenclature
XXIX
Greek
αl α il
α l ,max γv γ G γ Δ
δ δl ∂ ε η θ κ κ κl λ λ μlτ
part of γ vVol available to the velocity field l, local instantaneous volume fraction of the velocity field l, dimensionless the same as α l in the case of gas mixtures; in the case of mixtures consisting of liquid and macroscopic solid particles, the part of γ vVol available to the inert component i of the velocity field l, local instantaneous volume fraction of the inert component i of the velocity field l, dimensionless ≈ 0.62 , limit for the closest possible packing of particles, dimensionless the part of dVol available for the flow, volumetric porosity, dimensionless surface permeability, dimensionless directional surface permeability with components γ r , γ θ , γ z , dimensionless finite difference small deviation with respect to a given value = 1 for continuous field; = 0 for disperse field, dimensionless partial differential dissipation rate for kinetic energy from turbulent fluctuation, power irreversibly dissipated by the viscous forces due to turbulent fluctuations, W / kg dynamic viscosity, kg/(ms) θ -coordinate in the cylindrical or spherical coordinate systems, rad = 0 for Cartesian coordinates, = 1 for cylindrical coordinates isentropic exponent curvature of the surface of the velocity field l, m thermal conductivity, W/(mK) eigenvalue local volume-averaged mass transferred into the velocity field l per unit time and unit mixture flow volume, local volume-averaged instantaneous
(
mass source density of the velocity field l, kg / m3 s τ
(
)
)
μl
time average of μ l , kg / m s
μ wl
mass transport from exterior source into the velocity field l, kg / m3 s
τ
μil
3
(
)
local volume-averaged inert mass from species i transferred into the velocity field l per unit time and unit mixture flow volume, local volume-averaged
XXX
Nomenclature
instantaneous mass source density of the inert component i of the velocity
(
field l, kg / m3 s
)
(
time average of μ ilτ , kg / m3 s
μ il τ
μ iml
)
local volume-averaged instantaneous mass source density of the inert component i of the velocity field l due to mass transfer from field m,
(
kg / m3 s
)
(
τ time average of μ iml , kg / m3 s
μ iml τ
μ ilm
)
local volume-averaged instantaneous mass source density of the inert component i of the velocity field l due to mass transfer from field l into
(
) , kg / ( m s )
velocity field m, kg / m3 s τ
μ ilm
time average of μ ilm
ν
cinematic viscosity, m 2 / s
ν
t l
3
coefficient of turbulent cinematic viscosity, m 2 / s
ν ltn ξ ρ ρ ρl ρil
coefficient of turbulent particles diffusion, m 2 / s angle between nlσ and ΔVlm , rad density, kg/m3 instantaneous density, density; without indexes, mixture density, kg/m3 instantaneous field density, kg/m3 instantaneous inert component density of the velocity field l, kg/m3
ρl
l
intrinsic local volume-averaged phase density, kg/m3
(
( ρ w )23 ( ρ w )32
(ρ V ) τ
l
l
entrainment mass flow rate, kg / m 2 s
(
2
deposition mass flow rate, kg / m s le
)
) (
local intrinsic surface mass flow rate, kg / m 2 s
)
σ , σ 12 surface tension between phases 1 and 2, N/m τ time, s ϕ angle giving the projection of the position of the surface point in the plane normal to ΔVlm , rad
χ
mσ l
the product of the effective heat transfer coefficient and the interfacial
(
)
area density, W / m3 K . The subscript l denotes inside the velocity field l. The superscript mσ denotes location at the interface σ dividing field m from field l. The superscript is only used if the interfacial heat transfer is associated with mass transfer. If there is heat transfer only, the
Nomenclature XXXI
linearized interaction coefficient is assigned the subscript ml only, indicating the interface at which the heat transfer takes place. Subscripts c d lm w e l i
continuous disperse from l to m or l acting on m region "outside of the flow" entrances and exits for control volume Vol velocity field l, intrinsic field average inert components inside the field l, non-condensable gases in the gas field l = 1, or microscopic particles in water in field 2 or 3 i corresponding to the eigenvalue λi in Chapter 4 M non-inert component m mixture of entrained coolant and entrained melt debris that is in thermal and mechanical equilibrium behind the shock front ml from m into l iml from im into il max maximum number of points n inert component 0 at the beginning of the time step E entrainment coal coalescence sp splitting, fragmentation σ interface τ old time level τ + Δτ new time level * initial 0 reference conditions p,v,s at constant p,v,s, respectively L left R right 1 2 3 4 5
vapor or in front of the shock wave water or behind the shock wave melt entrained coolant behind the front – entrained coolant micro-particles after the thermal interaction – entrained melt
Superscripts ´ ' "
time fluctuation saturated steam saturated liquid
XXXII
"' A d e i imax L l le lσ m n n n+1 t vm
τ
Nomenclature
saturated solid phase air drag heterogeneous component (either gas or solid particles) of the velocity field maximum for the number of the components inside the velocity field lift intrinsic field average intrinsic surface average averaged over the surface of the sphere component normal old iteration new iteration turbulent, tangential virtual mass temporal, instantaneous averaging sign
Operators ∇⋅ ∇ ∇n
divergence gradient normal component of the gradient tangential component of the gradient surface gradient operator, 1/m
∇t ∇l
∇2
Laplacian local volume average l
local intrinsic volume average
le
local intrinsic surface average
Nomenclature required for coordinate transformations
( x, y, z ) coordinates
of a Cartesian, left oriented coordinate system (Euclidean
space). Another notation which is simultaneously used is xi
( i = 1, 2,3) :
x1 , x2 , x3
(ξ ,η , ζ ) coordinates of the curvilinear coordinate system called transformed coordinate system. Another notation which is simultaneously used is ξ i
( i = 1, 2,3) : ξ 1 , ξ 2 , ξ 3
Nomenclature XXXIII
the velocity of the curvilinear coordinate system
Vcs
Jacobian determinant or Jacobian of the coordinate transformation
g
x = f ( ξ , η , ζ ) , y = g (ξ , η , ζ ) , z = h (ξ , η , ζ )
elements of the Jacobian determinant
aij a
ij
elements of the determinant transferring the partial derivatives with respect to the transformed coordinates into partial derivatives with respect to the physical coordinates. The second superscript indicates the Cartesian components of the contravariant vectors
( a1 , a 2 , a3 )
covariant base vectors of the curvilinear coordinate system tangent vectors to the three curvilinear coordinate lines represented by (ξ , η , ζ )
(a , a , a ) 1
2
3
contravariant base vectors, normal to a coordinate surface on which the coordinates ξ , η and ζ are constant, respectively covariant metric tensor (symmetric)
g ij g
ij
contravariant metric tensor (symmetric)
(e , e , e ) 1
2
3
unit vectors normal to a coordinate surface on which the coordinates
ξ , η and ζ are constant, respectively = ai ⋅ V , contravariant components of the vector V = ai ⋅ V , covariant components of the vector V
i
V Vi
(γ
ξ
,γ η ,γ ζ
) permeabilities of coordinate surfaces on which the coordinates ξ , η and ζ are constant, respectively
Greek Α, α Β, β Γ, γ Δ, δ Ε, ε Ζ, ζ Η, η Θ, ϑ
Alpha Beta Gamma Delta Epsilon Zeta Eta Theta
Ι, ι Κ,κ Λ, λ Μ, μ Ν,ν Ξ ,ξ Ο, ο Π, π Ρ, ρ
Iota Kappa Lambda Mu Nu Xi Omikron Pi Rho
Σ, σ Τ, τ Φ, ϕ Χ, χ ϒ, υ Ψ, ψ Ω, ω
Sigma Tau Phi Chi Ypsilon Psi Omega
Introduction
Multi-phase flows are not only part of our natural environment such as rainy or snowy winds, tornadoes, typhoons, air and water pollution, volcanic activities etc., see Fig.1, but also are working processes in a variety of conventional and nuclear power plants, combustion engines, propulsion systems, flows inside the human body, oil and gas production and transport, chemical industry, biological industry, process technology in the metallurgical industry or in food production etc.
Fig. 1. The fascinating picture of the discovery start, the peace of universe, the tornado, the volcano, the flows in the human heart or even the “pure” water or the sky in the picture of Van Gogh are in fact different forms of multiphase flows
The list is by far not exhaustive. For instance everything to do with phase changes is associated with multi-phase flows. The industrial use of multi-phase systems requires methods for predicting their behavior. This explains the “explosion” of scientific publications in this field in the last 50 years. Some countries, such as Japan, have declared this field to be of strategic importance for future technological development. Probably the first known systematic study on two-phase flow was done during the Second World War by the Soviet scientist Teletov and published later in 1958 as “On the problem of fluid dynamics of two-phase mixtures”. Two books that appeared in Russia and the USA in 1969 by Mamaev et al. and by Wallis played an important role in educating a generation of scientists in this discipline including me. Both books contain valuable information mainly for steady state flows in pipes. Hewitt and Hall-Taylor published in 1974 “Annular two-phase flow”. The book also considers steady state pipe flows. The usefulness of the idea of a threefluid description of two-phase flows was clearly demonstrated on annular flows with entrainment and deposition. Ishii published in 1975 the book “Thermo-fluid
XXXVI
Introduction
dynamic theory of two-phase flow”, which contained a rigorous derivation of time-averaged conservation equations for the so called two-fluid separated and diffusion momentum equations models. This book founded the basics for new measurement methods appearing on the market later. The book was updated in 2006 by Ishii and Hibiki by including new information about the interfacial area density modeling in one-dimensional flows developed by the authors for several years. R. Nigmatulin published “Fundamentals of mechanics of heterogeneous media” in Russian in 1978. The book mainly considers one-dimensional two-phase flows. Interesting particular wave dynamics solutions are obtaining for specific sets of assumptions for dispersed systems. The book was extended mainly with mechanical interaction constitutive relations and translated into English in 1991. The next important book for two-phase steam-water flow in turbines was published by Deich and Philipoff in 1981 in Russian. Again mainly steady state, onedimensional flows are considered. Delhaye et al. published in the same year “Thermohydraulics of two-phase systems for industrial design and nuclear engineering”. The book contains the main ideas of local volume averaging, and considers mainly many steady state one-dimensional flows. One year later, in 1982, Hetsroni edited the “Handbook of multi phase systems” containing the state of the art of constitutive interfacial relationships for practical use. The book is still a valuable source of empirical information for different disciplines dealing with multi-phase flows. Crowe 2006 edited in 2006 the “Multiphase flow handbook” containing updated state of the art of constitutive interfacial relationships for practical use. In the monograph “Interfacial transport phenomena” published by Slattery in 1990 complete, rigorous derivations of the local volume-averaged twofluid conservation equations are presented together with a variety of aspects of the fundamentals of the interfacial processes based on his long years of work. Slattery’s first edition appeared in 1978. Some aspects of the heat and mass transfer theory of two-phase flow are now included in modern text books such us “Thermodynamics” by Baer (1996) and “Technical thermodynamics” by Stephan and Mayinger (1998). It is noticeable that none of the above mentioned books is devoted in particular to numerical methods of solution of the fundamental systems of partial differential equations describing multi-phase flows. Analytical methods still do not exist. I published in 1986 the book “Transient two-phase flows” with Springer-Verlag in German, discussing several engineering methods and practical examples for integrating systems of partial differential equations describing two- and three-fluid flows in pipes. Since 1984 I have worked intensively on creating numerical algorithms for describing complicated multi-phase multi-component flows in pipe networks and complex three-dimensional geometries mainly for nuclear safety applications. Note that the mathematical description of multi-dimensional two- and multi-phase flows is a scientific discipline with considerable activity in the last 30 years. In addition thousands of scientists have collected for years experimental information in this field. But there is still a lack of a systematic presentation of the theory and practice of numerical multi-phase fluid dynamics. This book is intended to fill this gap.
Introduction
XXXVII
Numerical multi-phase fluid dynamics is the science of the derivation and the numerical integration of the conservation equations reflecting the mass momentum and energy conservation for multi-phase processes in nature and technology at different scales in time and space. The emphasis of this book is on the generic links within computational predictive models between • • • •
fundamentals, numerical methods, empirical information for the constitutive interfacial phenomena, and comparison with experimental data at different levels of complexity.
The reader will realize how strong the mutual influence of the four model constituencies is. There are still many attempts to attack these problems using singlephase fluid mechanics by simply extending existing single-phase computer codes with additional fields and linking with differential terms outside of the code without increasing the strength of the feedback in the numerical integration methods. The success of this approach in describing low concentration suspensions and dispersed systems without strong thermal interactions should not confuse the engineer about the real limitations of this method. This monograph can be considered also as a handbook on numerical modeling of three strongly interacting fluids with dynamic fragmentation and coalescence representing multi-phase multi-component systems. Some aspects of the author's ideas, such us the three-fluid entropy concept with dynamic fragmentation and coalescence for describing multi-phase, multi-component flows by local volumeaveraged and time-averaged conservation equations, have been published previously in separate papers but are collected here in a single context for the first time. An important contribution of this book to the state of the art is also the rigorous thermodynamic treatment of multi-phase systems, consisting of different mixtures. It is also the first time of publishing the basics of the boundary fitted description of multi-phase flows and an appropriate numerical method for integrating them with proven convergence. It is well known in engineering practice that “the devil is hidden in the details”. This book gives many hints and details on how to design computational methods for multi-phase flow analysis and demonstrates the power of the method in the attached compact disc and in the last chapter in Volume 2 by presenting successful comparisons between predictions and experimental data or analytical benchmarks for a class of problems with a complexity not known in the multi-phase literature up to now. It starts with the single-phase U-tube problem and ends with explosive interaction between molten melt and cold water in complicated 3D geometry and condensation shocks in complicated pipe networks containing acoustically interacting valves and other components. Volume 3 is devoted to selected subjects in the multiphase fluid dynamics that are very important for practical applications but could not find place in the first two volumes. The state of the art of the turbulence modeling in multiphase flows is presented. As introduction, some basics of the single-phase boundary layer theory including some important scales and flow oscillation characteristics in pipes and rod bundles are presented. Then the scales characterizing the dispersed flow
XXXVIII
Introduction
systems are presented. The description of the turbulence is provided at different level of complexity: simple algebraic models for eddy viscosity, algebraic models based on the Boussinesq hypothesis, modification of the boundary layer share due to modification of the bulk turbulence, modification of the boundary layer share due to nucleate boiling. Then the role of the following forces on the matematical description of turbulent flows is discussed: the lift force, the lubrication force in the wall boundary layer, and the dispersion force. A pragmatic generalization of the k-eps models for continuous velocity field is proposed containing flows in large volumes and flows in porous structures. Method of haw to derive source and sinks terms for multiphase k-eps models is presented. A set of 13 single- and two phase benchmarks for verification of k-eps models in system computer codes are provided and reproduced with the IVA computer code as an example of the application of the theory. This methodology is intendet to help other engineers and scientists to introduce this technology step-by-step in their own engineering practice. In many practical application gases are solved in liquids under given conditions, released under other conditions and therefore affecting technical processes for good of for bad. There is almost no systematic description of this subject in the literature. That is why I decided to collect in Volume 3 useful information on the solubility of oxygen, nitrogen, hydrogen and carbon dioxide in water under large interval of pressures and temperatures, provide appropriate mathematical approximation functions and validate them. In addition methods for computation of the diffusion coefficients are described. With this information solution and dissolution dynamics in multiphase fluid flows can be analyzed. For this purpose the non-equilibrium absorption and release on bubble, droplet and film surfaces under different conditions is mathematically described. In order to allow the application of the theory from all the three Volumes also to processes in combustion engines a systematic set of internally consistent state equations for diesel fuel gas and liquid valid in broad range of changing pressure and temperature are provided also in Volume 3. After systematically rewriting the information collected in these three Volumes into a universal computer code named IVA, I was capable to successfully simulate very complex processes some of them illustrated in Fig. 2. In 1984 no body, even my closest friends, believed that this will be one day possible. Now it is reality. I very much hope that this experience will be a productive source in the hands of knowledgeable engineers and scientists.
Introduction
XXXIX
Fig. 2. Examples of multiphase flows in the nuclear technology. See http://www.herzovision.de/kolev-nikolay/
Erlangen, Autumn 2007
Nikolay Ivanov Kolev
XL
Introduction
Tghgtgpegu" Baer HD (1996) Thermodynamik, Springer, Berlin Heidelberg New York Crowe CT ed. (2006) Multiphase flow handbook, Taylor&Francis, Boca Raton, London, New York Deich ME, Philipoff GA (1981) Gas dynamics of two phase flows. Energoisdat, Moscow (in Russian) Delhaye JM, Giot M, Reithmuller ML (1981) Thermohydraulics of two-phase systems for industrial design and nuclear engineering, Hemisphere, New York, McGraw Hill, New York Hetstroni G (1982) Handbook of multi phase systems. Hemisphere, Washington, McGrawHill, New York Hewitt GF and Hall-Taylor NS (1974) Annular two-phase flow, Pergamon, Oxford Ishii M (1975) Thermo-fluid dynamic theory of two-phase flow, Eyrolles, Paris Ishii M and Hibiki T (2006) Thermo-fluid dynamics of two-phase flow, Springer, NY Mamaev WA, Odicharia GS, Semeonov NI, Tociging AA (1969) Gidrodinamika gasogidkostnych smesey w trubach, Moskva Nigmatulin RI (1978) Fundamentals of mechanics of heterogeneous media, Nauka, Moscow, 336 pp (in Russian) Nigmatulin RI (1991) Dynamics of multi-phase media, revised and augmented edition, Hemisphere, New York Slattery JC (1990) Interfacial transport phenomena, Springer, Berlin Heidelberg New York Stephan K and Mayinger F (1998) Technische Thermodynamik, Bd.1, Springer, 15. Auflage Teletov SG (1958) On the problem of fluid dynamics of two-phase mixtures, I. Hydrodynamic and energy equations, Bulletin of the Moscow University, no 2 p 15 Wallis GB (1969) One-dimensional two-phase flow, McGraw-Hill, New York
3"Ocuu"eqpugtxcvkqp"
“...all changes in the nature happen so that the mass lost by one body is added to other...”. 1748, M. Lomonosov’s letter to L. Euler
303"Kpvtqfwevkqp" The creation of computer codes for modeling multi-phase flows in industrial facilities is very complicated, time-consuming and expensive. That is why the fundamentals on which such codes are based are subject to continuous review in order to incorporate the state of the art of knowledge into the current version of the code in question. One of the important elements of the codes is the system of partial differential equations governing the flow. The understanding of each particular term in these equations is very important for the application. From the large number of formulations of the conservation equations for multi-phase flows, see Slattery (1990), Hetstrony (1982), Delhaye et al. (1981), Ishii (1975) and the references given there, the local volume averaging as founded by Anderson and Jackson (1967), Slattery (1967), Whitaker (1967) is selected because it is rigorous and mathematical elegant. The heterogeneous porous-media formulation introduced by Gentry, et al. (1966), commented by Hirt (1993), and used by Sha et al. (1984), is then implanted into the formalism as a geometrical skeleton because of its practical usefulness. For technical structures, the introduction of the local volume porosity and directional permeability is a convenient formalism to describe the real distances between the flow volumes. Geometry simplifications that lead to distortion of the modeled acoustic process characteristics of the systems are not necessary. Beyond these concepts, I include inert components in each velocity field, and introduce the concept of dynamic particle fragmentation and coalescence. Then I perform subsequent time averaging. The link between kinetic generation of particles and the mass source terms is specified. The link between mass source terms and the change in bubble/droplet size due to evaporation or condensation is presented for local volume- and time- averaged source terms. The concept of mono-dispersity is discussed and a method is proposed for computation of the disappearance of particles due to evaporation and condensation.
2
1 Mass conservation
Starting this work in 1983 for creation of the foundations of the IVA computer code I believed, and now after 23 years, I am convinced, that this concept is unique and that it is the most effective concept for deriving multi-phase equations applicable to flows in complicated technical facilities. Here I will present the derivation of the mass conservation equation for single component inside the field, for the mixture constituting the field. Then I will present the derivation of the particle number density equations. This Chapter is an improved version of the work published in Kolev (1994b).
304"Dcuke"fghkpkvkqpu" Consider the idealized flow patterns of the three-phase flows shown in Fig. 1.1. What do all these flow patterns have in common? How can they be described by a single mathematical formalism that allows transition from any flow pattern to any other flow pattern? How can the complicated interactions between the participating components be described as simply as possible? In the discussion that follows a number of basic definitions are introduced or reiterated so as to permit successful treatment of such flows using a unified algorithm. Zemansky (1968) p. 566 defines “a phase as a system or a portion of the system composed of any number of chemical constituents satisfying the requirements (a) that it is homogeneous and (b) that it has a definite boundary”. In this sense, consideration of a multi-phase flow system assumes the consistency described below: 1) A homogeneous gas phase having a definite boundary. The gas phase can be continuous (non-structured) or dispersed (structured). The gas phase in the gas/droplet flow is an example of a non-structured gas phase. The bubbles in the bubble/liquid flow are an example of a structured gas phase. Besides the steam, the gas phase contains i1max − 1 groups of species of inert components. Inert means that these components do not experiences changes in the state of aggregate during the process of consideration. The gas components occupy the entire gas volume in accordance with the Dalton’s law, and therefore possess the definite boundary of the gas phase it selves. The gas phase velocity field will be denoted as no.1 and l = 1 assigned to this. Velocity field no.1 has the temperature T1 . 2) A mixture of liquid water and a number of species i2 max − 1 of microscopic solid particles. The particles are homogeneous and have a definite boundary surrounded by liquid only. The water has its own definite outer boundary. An example of such a mixture is a liquid film carrying i2 max − 1 groups of radioactive solid species. This means that the mixture of water with i2 max − 1 groups of species is a mixture of i2 max phases. The mixture velocity field defined here will be denoted as no. 2. Velocity field no. 2 can likewise be non-structured or structured. Velocity field no. 2 has the temperature T2 .
1.2 Basic definitions 2
1
7
6
4
3
8
5
9
3
Gas
Water + microscopic solid
10
particles 11
12
14
13
15
Liquid metal
16
17
18
19
20 Pool flow
22
21
Pool + channel flow
23 24
Liquid + solid
Solid
Bubbles
microscopic solid particles
particles
26
flow
Water _ droplets +
Water _ film + microscopic solid
25
Channel
27
28
Fig. 1.1 Multi-phase flow patterns
3) A mixture of the type of velocity field no. 2 of liquid with i3max − 1 groups of species of microscopic solid particles. The mixture can be non-structured or structured; this is denoted as velocity field no. 3. Velocity field no. 3 has the temperature T3 . In the event that only one inert component occupies velocity field no. 3, this will be allowed to be macroscopic, either a liquid, a liquid-solid mixture in homogenous equilibrium or solid particles only. This allows versatile application of the concept.
4
1 Mass conservation
The entire flow under consideration consists of i2 max + i3max + 1 phases and is conditionally divided into three velocity fields. In order to avoid the need to write indices to the indices, il will be written in place of il . An example demonstrating the necessity for the use of three velocity fields is the modeling of a mixture consisting of gas, water, and a liquid metal, whose densities have the approximate ratios 1: 103 :104 . In a transient, these three fields will have considerably differing velocities and temperatures. Turning our attention again to the flow patterns depicted in Fig. 1.1, and keeping in mind that in reality these change their characteristic sizes chaotically, it is obvious that it is not possible to determine the details of the thermodynamic and flow parameters for each component of each velocity field. Consequently, some type of averaging must be implemented. Following Anderson and Jackson (1967), Slattery (1967, 1990), Whitaker (1967), a control volume is allocated to every point in space, the thermodynamic and flow properties averaged for each velocity field are assigned to the center of the control volume. Since there is a control volume associated with every point in space, a field of average values can be generated for all thermodynamic and flow properties. The field of the average properties is therefore smooth, and space derivatives of these averaged properties exist. Consider the control volume Vol occupied by a non-movable structure in addition to the three velocity fields, see Figs. 1.2 and 1.3. While the individual lfield volumes Voll may be functions of time and space, the control volume Vol is not. Velocity field l is taken to have the characteristic length Dl (e.g. bubble size, droplet size), which is much larger than the molecular free path, Fig. 1.3. The size of the computational region of interest here is much larger than the size of the local structure Dl and larger than the spatial changes in the flow parameters of interest. The choice of the size of the control volume, which is of order of
(1 − γ v )Vol
Vol
m n lσ
l
lmax
∑ Vol
Vmτ
Flσ
Fle
l
l =1
w, structures
θ
z
Vlτ
r
Fig. 1.2 Control volume for definition of the mass conservation equation, partially occupied by structure and two different velocity fields
1.2 Basic definitions
5
Vol 1/3
Dl
Vol Δz
Δr
s. c.r.
Fig. 1.3 Comparison of possible scales of local volume averaging, scale of the measuring devices, scale of the computational region (s.c.r.) and the global flow dimensions
Vol1/ 3 has a major impact on the meaning of the averaged values. From the various approaches possible, there are two which are meaningful.
i) The size of the control volume is larger than the characteristic configuration length Dl of the field l, i.e., molecular free path 0 . For simplicity, any averaging signs are omitted. nl takes values between zero and approximately 1012 per cubic meter (1012 m-3 is presumably the upper limit for nucleation during evaporation or condensation). nl = 0 for an existing velocity field here will be associated with a continuous fluid. 0 < nlVolcel < 1 , where Volcel is the volume under consideration, can be interpreted as continuous fluid (jet or free surface) with excited interface. nlVolcel → 0 means a stable surface, and nlVolcel → 1 means an unstable interface. nlVolcel > 1 means that there is no longer a continuous velocity field l. Here −ν ltn / Scltn is the diffusive flux for particle number density resulting from the fact that the particles are in random motion in the presence of a particle density gradient. In accordance with Batchelor (1970), −
ν ltn Sc
1/ 2
tn
≈ const Dl ( Vl′ ⋅ Vl′)
= const Dl ( ΔVlm ⋅ ΔVlm )
1/ 2
⎡⎣ H (α l ) ⎤⎦ , (1.111) 1/ 2
where the constant is approximately unity, H (α l ) ≈
αl α l ,max
⎛ αl ⎜⎜1 − ⎝ α l ,max
⎞ ⎟⎟ , ⎠
(1.112)
and α l ,max ≈ 0.62 is the limit for the closest possible packing of particles. ( Vl′ ⋅ Vl′) is the mean square of the velocity fluctuations and ( ΔVlm ⋅ ΔVlm )
1/ 2
is the magni-
tude of the relative velocity. This choice meets the requirements for disappearance of velocity fluctuations at the limit α l → 0 and at the limit for the closest possible packing α l → α l ,max . For the locally mono-disperse system there is a unique relationship between volume, concentration, particle number density and the local particle length scale or the interfacial area density ad (interfacial area Fd divided by the mixture flow volume
lmax
∑ Vol l =1
ad = Fd
l
),
lmax
∑ Vol l =1
l
= nd Fd ,single particle = α d Fd ,single particle Vold ,single particle 1/ 3
= αd
6 ⎛π ⎞ = 6 ⎜ nd ⎟ α d2 / 3. Dd ⎝6 ⎠
(1.113)
1.10 Conservation equations for the number density of particles
27
Therefore the use of one of the variables nd or Dd or ad is equivalent and only the simplicity of the conservation equation dictates to prefer nd . The production terms in the right-hand side of Eq. (1.109) can be classified as kinetic and non-kinetic terms. The kinetic terms are nl ,kin = f lw
imax 4 ′′ + α m fim ∑ nim + α m ( nl ,h + nl ,dis ). nlw Dhy i =1
(1.114)
In the case of bubble flow, l = 1, m = 2. For this case, the following then apply: 4 ′′ is the number of bubbles generated at the wall per unit time and unit flw nlw Dhy volume of the flow (wall cavity nucleation rate). flw is the frequency of bubble ′′ is the number of actigeneration at one activated seed on the channel wall. nlw vated seeds per unit area of the wall. imax
The term α m f im ∑ nim gives the number of bubbles generated from the solid i =1
particles homogeneously mixed with the second velocity field per unit time and unit volume of the flow. f im is the frequency of bubbles generated from one activated seed on the particle belonging to the second velocity field. nl ,h + nl ,dis is the bulk liquid nucleation rate consisting of the number of bubbles generated by homogeneous nucleation in the second velocity field per unit time and unit volume of the flow, nl ,h , and the number of bubbles generated from the dissolved gases in the second velocity field per unit time and unit volume of the flow, nl ,dis . The non-kinetic mechanical terms are nl ,coal = nl f l ,coal / 2 ,
(1.115)
this denoting the number of the bubbles that disappear due to coalescence per unit time and unit volume of the flow, and nl ,sp = nl f l ,sp ,
(1.116)
which is the number of the bubbles arising due to hydrodynamic bubble disintegration per unit time and unit volume of the flow. f l ,coal is the coalescence frequency of two colliding particles. f l ,sp is the fragmentation frequency of single particle. The coalescence frequency of a single particle is defined as a product of the collision frequency per single particle f d ,col and the coalescence probability of two colliding particles Pd ,coal . Different causes for the collisions are associated with different collision frequencies f ds,col ,
f dno,col and
f do,col . Similarly, different causes for collisions are associated with different
28
1 Mass conservation
coalescence probabilities Pdno,coal , Pdo,coal . Pdo,coal is the probability of oscillatory coalescence (due to turbulent oscillations). Pdno,coal is the probability of nonoscillatory coalescence (due to non-uniform velocity field in space). Thus the final expression takes the form f d ,coal = ( f ds,col + f dno,col ) Pdno,coal + f do,col Pdo,coal .
(1.117)
f do,col is the frequency of collision due to turbulence fluctuation of the particles. It
depends on the fluctuation component of the velocity of the particles, Vd′ , f do,col = f ( ΔVd′ , α d ,...,...) .
(1.118)
The superscript o is used to remember that this is an oscillatory frequency. f ds,col and f dno,col are the frequencies of collision due to the convective motion. The splitting in two components is due to the use of the concept of mono-dispersity. The computational collisions, f dno,col , are caused only by the mean relative velocity between the field with averaged particle size and the surrounding continuum ΔVddno . Correct mathematical averaging of the velocity difference gives an additional component ΔVdds associated with f ds,col . f ds,col is zero for real mono-disperse systems. Thus f dno,col = f ( ΔVddno , α d ,...,...) ,
(1.119)
f ds,col = f ( ΔVdds , α d ,...,...) .
(1.120)
and
The superscript no is used to remember that this is a non-oscillatory frequency. The superscript s is used to remember that this frequency is associated with a spectrum of a particle sizes. In a similar way, the source terms for the conservation equation for droplet number density, l = 3 are defined. For channel flow it is assumed here that the second velocity field is continuous and the third disperse. In this case, besides the fragmentation and coalescence that also exist in pool flow, the entrainment and deposition will influence not only the mass but also the particle number density balance
μ 23 − μ32 = n23 − n32 =
4 1/ 2 (1 − α 2 ) ⎡⎣( ρ w )23 − ( ρ w )32 ⎤⎦ , Dh
6 4 1/ 2 (1 − α 2 ) ⎡⎣( ρ w )23 π Dh
(1.121)
( ρ D ) − ( ρ w) ( ρ D )⎤⎦ , 2
3 3E
32
3
3 3
(1.122)
1.10 Conservation equations for the number density of particles
29
where ( ρ w )23 and ( ρ w )32 are entrainment and deposition mass flow rate and D3 E is the diameter of the entrained droplets. The large variety of phenomena leading to fragmentation and coalescence are discussed by the author in Kolev (1993a), and more recently Kolev (1998). In Volume II of this monograph the reader will find the current status of empirical knowledge in this field and the way in which this can be used to compute the fragmentation and coalescence production rates. It is appropriate now to highlight an interesting potential of this concept. Multiplying nl with
3
∑ Vol l =1
l
gives the number of particles in a single control volume or
in a single discretization volume during numerical integration. If there is more than one particle in the cell, 3
nl ∑ Voll > 1 ,
(1.123)
l =1
the field is dispersed. Consequently, the mechanisms governing fragmentation and coalescence are the mechanisms for dispersed field l surrounded by m, e.g., droplets or bubble fragmentation and coalescence. If 3
nl ∑ Voll ≤ 1 ,
(1.124)
l =1
the field is continuous. This means that the mechanisms controlling potential fragmentation are the mechanisms known for continuous field l surrounded by the field m. An example is jet fragmentation, where both participating fields at the jet region are continuous. In this case there is no coalescence. Thus, natural transition from continuum to disperse and vice versa can be modeled. In addition, a very important memory effect of the multi-phase structure is taken into account in this modeling approach. An example for the use of this approach was given in Kolev (1993b),, for interaction and fragmentation of gas, molten metal and water during a transient. Equation (1.109) without diffusion and non-kinetic production terms was successfully used by Kocamustafaogulari and Ishii (1983) for the modeling of single-component boiling systems. A comparison of non-equilibrium model predictions with experimental data for flashing in Laval nozzles performed by the author (1985b) showed that an additional differential equation for description of the particle number density is necessary to obtain a more accurate prediction than the widespread approach of assuming an almost arbitrary number density within the range of 109 to 1013cm-3. Riznic and Ishii (1989) applied the Kocamustafaogulari and Ishii approach with success to the modeling of singlecomponent flushing systems. Deich and Philippoff (1981) analyzed the pressure and temperature distribution inside the eddies of subcooled steam and came to the conclusion that eddies with appropriate dimensions serve as a nuclei for condensation, this providing further substantiation for use of the method discussed above.
30
1 Mass conservation
The derived particle number density conservation equation can be used to compute the surface energy associated with the interface. Multiplying the bubble number density, nl , by the surface energy of a single bubble, π D12σ 12 , the surface energy per unit mixture volume is then obtained. The conservation equation for this energy is similar to Eq. (1.109). It is interesting to note that the surface energy is transported by the convection and diffusion of the discrete velocity fields but that the terms supplying the change in this energy (kinetic origination, collision, splitting) result from the energy of the surrounding continuous liquid.
3033"Kornkecvkqp"qh"vjg"cuuworvkqp"qh"oqpq/fkurgtukv{" kp"c"egnn" 303303"Rctvkeng"uk|g"urgevtwo"cpf"cxgtcikpi"
The result of the initial, boundary conditions, convection and local coalescence, fragmentation entrainment and deposition is a spectrum of particle size. In this Section we will demonstrate what influence the replacement of the spectrum of particle sizes with a single particle size may have on heat and mass transfer processes and how it can be incorporate approximately in the analysis. For the illustration of this we use the observed in 1931 Nukiama - Tanasawa distribution P ( Ddi ) ⎛ D ⎞ −2 D = 4 ⎜ di ⎟ e ( di Dd′ ⎝ Dd′ ⎠ 2
Dd′ )
.
(1.125)
Here P ( Ddi ) is the probability that a particle size is between Ddi and Ddi + δ Ddi . Dd′ is the most probable particle size, the size where the probability distribution function has its maximum value. The particle sizes may take values between zero and a maximum value
0 < Ddi < Dd ,max .
(1.126)
Thus, if we know Dd′ and Dd ,max , the particle distribution is uniquely characterized. MacVean (see Wallis (1969)) found that a great deal of data could be correlated by assuming that Dd′ = Dd /2,
(1.127)
where Dd is the volume-averaged particle size. The relationship between Dd and the maximum particle size, Dd ,max , is reported as Dd ,max ≈ (2.04 to 3.13) Dd
(1.128)
1.11 Implication of the assumption of mono-dispersity in a cell
31
(see Pilch et al. (1981) and Kataoka et al. (1983), among others. Other distributions give slightly different results e.g. Kolomentzev and Dushkin (1985) P ( Ddi ) ⎛ D ⎞ −4 D = 8 ⎜ di ⎟ e ( di Dd′ ⎝ Dd′ ⎠ 2
2 Dd′ )
(1.129)
came to Dd ,max ≈ 1.33 Dd . Other distributions are reported by Rosin and Rammler (1933), Griffith (1943) and Mugele and Evans (1951). In contrast to equation (1.113) for mono-disperse spherical particles the interfacial area density in the general case is ad = Fd =
lmax
∑ Vol l =1
l
= nd Fd ,single particle = α d Fd ,single particle Vold ,single particle
6α d 6α d Dd = . Ddsm Dd Ddsm
(1.130)
Here sm d
D
⎡ Dd ,max 3 = ⎢ ∫ Dd P ( Dd ) dDd ⎢⎣ 0
Dd ,max
∫ 0
⎤ Dd2 P ( Dd ) dDd ⎥ ⎥⎦
(1.131)
is the so called Sauter mean diameter, see Sauter (1929). A droplet with diameter equal to the Sauter mean diameter has the same surface to volume ratio as the entire spray. Using the Nukiama-Tanasawa distribution, Eq. (1.125), and Eqs. (1.127) and (1.128), after evaluating Eq. (1.131) analytically we obtain for the Sauter mean diameter Ddsm ≈ (1.154 to 1.238) Dd
( Ddsm < 1.25Dd ),
or Dd Ddsm ≈ 0.867 to 0.807 > 0.8,
(1.132)
which is perfectly confirmed by the measurements summarized by Faeth (1995) Ddsm / Dd ≈ 1.2 . This is a very useful result, allowing use of the volume fraction α d and the particle number density nd in cases where the assumption of monodispersity does not hold. 303304"Ewvvkpi"qh"vjg"nqygt"rctv"qh"vjg"urgevtwo"fwg"vq"ocuu"vtcpuhgt"
Use of the particle number density conservation equations for each of the three velocity fields as already implemented in the IVA computer code series since 1985 is an extremely practical method of modeling the scale of the field. A number of important features of the concept of mono-dispersity will now be discussed.
32
1 Mass conservation
The time-averaged source terms associated with the change in the aggregate state for the mass conservation equations are
μl =
1 Δτ
Δτ
∫ 0
d ⎛ π 3⎞ ⎡ ⎛ π 3 ⎞⎤ ⎜ nl ρl Dl ⎟dτ = ⎢ Δ ⎜ nl ρl Dl ⎟ ⎥ Δτ dτ ⎝ 6 6 ⎠ ⎠⎦ ⎣ ⎝
imax ⎤ π⎡ 4 3 ′′ + Dldm = ρl 0 ⎢ Dld3 flw nlw α m fim ∑ nim + α m ( Dl3,h nl ,h + Dl3,dis nl ,dis ) ⎥ 6 ⎣⎢ Dhy i =1 ⎦⎥
+ α l 0 ρl 0 ⎡⎣ ρl Dl3 / ( ρl 0 Dl30 ) − 1⎤⎦ Δτ .
(1.133)
Here ρl 0 , nl 0 and Dl 0 are density, particle number density, and particle size at the beginning of the time step Δτ . Dld is the size with which the bubbles are produced after one nucleation cycle on the solid structure, i.e., the particle departure diameter. Dldm is the size with which the bubbles are produced after one nucleation cycle on the inert solid particles of the field m = 2. If the nucleation takes place in the bulk of the donor field, the size is equal to the smallest stable size Dl ,h , or in case of nucleation on dissolved gases, Dl ,dis . More information for modeling of particulate processes is provided in Volume II of this monograph. The first term in the above equation gives the time and local volume-averaged mass production due to nucleation; the second term gives the time and local volume average of mass production due to mass change for existing particles. It is important to note that the second term describes changes in the particle size due to evaporation and condensation. In accordance with the concept of monodispersity, the mass changes leading to a decrease in the local volume and time average particle size also cause the disappearance of those particles in the spectrum having sizes smaller than the size change. This statement, together with the assumption made for the form of the spectrum, provides the basics for computation of the averaged number of disappearing particles. Suppose the particle size distribution obeys the Nukiama - Tanasawa (1931) law P ( Dl ) = 4 ( Dl / Dl′ ) exp ⎡⎣ −2 ( Dl / Dl′ )⎤⎦ . 2
(1.134)
Suppose the form of the distribution remains unchanged during the time step. Having in mind that μ lm = −α l 0 ρl 0 ⎡⎣ ρl Dl3 / ( ρl 0 Dl30 ) − 1⎤⎦ / Δτ the volume-averaged diameter changes within this time by ⎧⎪ ⎡⎛ μ Δτ ⎞ ρ ⎤1/ 3 ⎫⎪ ΔDl = Dl 0 − Dl = Dl 0 (1 − Dl / Dl 0 ) = Dl 0 ⎨1 − ⎢⎜ 1 − lm ⎟ l 0 ⎥ ⎬ α l 0 ρl 0 ⎠ ρ l ⎦ ⎪ ⎩⎪ ⎣⎝ ⎭ ⎧⎪ ⎡⎛ μ Δτ ⎞ ⎤1/ 3 ⎫⎪ ≈ Dl 0 ⎨1 − ⎢⎜ 1 − lm ⎟ ⎥ ⎬ . ⎪⎩ ⎣⎝ α l 0 ρl 0 ⎠ ⎦ ⎪⎭
(1.135)
1.11 Implication of the assumption of mono-dispersity in a cell
33
and all particles having sizes 0 ≤ Dl ≤ 2ΔDl , namely, 2 ΔDl
nl
∫ P ( D ) dD
0
l
l
⎧⎪ ⎛ ΔDl ⎞ ⎡ ΔDl ⎛ ΔDl ⎞ ⎤ ⎫⎪ = 2nl ⎨1 − exp ⎜ −4 1+ 2 ⎢1 + 4 ⎟ ⎜ ⎟ ⎥ ⎬ , (1.136) Dl′ ⎠ ⎣ Dl′ ⎝ Dl′ ⎠ ⎦ ⎪⎭ ⎝ ⎪⎩
disappear. The averaged particle sink per unit time and unit mixture volume is consequently ⎛ n nl , spectrum _ cut = ⎜ l ⎝ Δτ
⎛ ΔDl ⎞ ⎧⎪ ⎟ 2 ⎨1 − exp ⎜ −4 D ′ ⎠ ⎩⎪ l ⎝
⎞⎡ ΔDl ⎛ ΔDl ⎟ ⎢1 + 4 ⎜1 + 2 Dl′ ⎝ Dl′ ⎠ ⎢⎣
⎞ ⎤ ⎫⎪ ⎟ ⎥ ⎬ , (1.137) ⎠ ⎥⎦ ⎭⎪
for ΔDl ≥ Dl′ . The latter condition means that the averaged particle size is ΔDl , with particle size ranging approximately within the interval 2ΔDl . It was found that for shock condensation of bubbles the inclusion of Eq. (1.137) for prediction of a reduction in the number density for disappearance of steam volume fractions is very important, Kolev (1993d, 1993e, 1993f, 1996, 1999). The same is true for the disappearance of water droplets in highly superheated gases due to evaporation. The latter can be the case in steam explosion analysis where gas is superheated due to previous contact with melt, with this followed by water droplet entry into the superheated gas regions. Combustion processes and spray cooling of hot gas jets are further examples. Equation (1.137) can be written in the following form nl ,spectrum _ cut = nl fl ,spectrum _ cut ,
(1.138)
where f l ,spectrum _ cut =
2 Δτ
⎛ ΔDl ⎪⎧ ⎨1 − exp ⎜ −4 Dl′ ⎝ ⎪⎩
⎞⎡ ΔDl ⎛ ΔDl ⎟ ⎢1 + 4 ⎜1 + 2 Dl′ ⎝ Dl′ ⎠⎣
⎞ ⎤ ⎪⎫ ⎟⎥ ⎬ , ⎠ ⎦ ⎪⎭
(1.139)
is the frequency of disappearance of particles due to the spectrum cutting. 303305"Vjg"ghhgev"qh"vjg"cxgtcikpi"qp"vjg"ghhgevkxg"xgnqekv{"fkhhgtgpeg"
Consider i groups of particles with different diameters having a concentration per unit volume called particle number density ndi = nd P ( Ddi )
i = 1, I
(1.140)
that is dependent on the diameter. Here P ( Ddi ) is the probability of a particle to have a size between Ddi and Ddi + δ Ddi . We assume that the continuous velocity
34
1 Mass conservation
field possesses a constant velocity wc . The difference of the continuum and the particle velocity is approximately described by the following equations
(1 + bd ccdvm )
∂Δwcd ⎛ 1 1 ⎞ 3 ccdd +⎜ − ∇ p + b Δwcd Δwcd = 0 , ⎟ d ∂τ 4 Dd ⎝ ρc ρd ⎠
(1 + b c ) ∂Δ∂τw vm di cdi
cdi
(1.141)
d ⎛ 1 1 ⎞ 3 ccdi +⎜ − Δwcdi Δwcdi = 0 , ⎟ ∇p + bdi 4 Ddi ⎝ ρc ρ d ⎠
(1.142)
where bd = (α c ρ c + α d ρ d ) / ( ρ c ρ d ) , and bdi = (α c ρ c + α di ρ d ) / ( ρ c ρ d ) . In this case we see that each particular group possesses a different relative velocity with respect to the continuum and consequently the groups are moving relatively to each other. Therefore collision caused by particle-particle relative velocity in real nature is an inevitable phenomenon. For steady flow we have from Eqs. (1.141) and (1.142) 1/ 2
⎛ b cd D ⎞ Δwcdi = Δwcd ⎜ d dd di ⎟ ⎝ bdi ccdi Dd ⎠
= Δwcd f i .
(1.143)
Obviously there is a resulting average velocity 1 nd
ΔVdd =
∑n
di
Δwcd − Δwcdi = Δwcd
i
1 nd
∑n
di
1 − f i = Δwcd
i
∑ P (D ) 1− f di
i
,
i
(1.144) which is the 1 nd
∑n
di
1 − fi
(1.145)
i
part of the average relative velocity of the droplets with respect to the continuum. For mono-dispersed particles f1 = f 2 = ... = f i = 1
(1.146)
ΔVdd = 0 .
(1.147)
and
The expression 1 nd
∑ ndi 1 − fi = i
Dd ,max / Dd
∫
P ( Ddi ) 1 − f i d ( Ddi / Dd ) ,
(1.148)
0
can be numerically estimated using known distribution functions e.g. the Nukiama Tanasava distribution (1939) as given by Eq. (1.125). Furthermore
1.13 Final remarks and conclusions
35
1/ 2
⎛ b * + 1 cdd Ddi ⎞ fi = ⎜ ⎟ d ⎝ b * + α di / α d ccdi Dd ⎠
,
(1.149)
where b* = α c ρc / α d ρ d can be expressed as function of Ddi / Dd having in mind that α di = (π Ddi3 / 6 ) ndi and α d = (π Dd3 / 6 ) nd and using Eq. (1.140),
α di / α d = ( Ddi / Dd ) P ( Ddi ) . 3
(1.150)
3034"Uvtcvkhkgf"uvtwevwtg" If the selected scale of discretization δ is smaller than the characteristic length of the fields,
δ κl < 1,
(1.151)
the local control volume contains a surface dividing for instance field l from field m. The unit vector nl pointing outwards from the field l is an important local characteristic of this surface. It is defined from the spatial distribution of the volume fraction of the field l as follows nl = −
∇ (α lγ )
∇ (α lγ )
.
(1.152)
The curvature of the surface is conveniently computed following the derivation of Brackbill, et al. (1992),
κ l = ∇ ⋅ nl ,
(1.153)
and is defined positive if the center of the curvature is in field l.
3035"Hkpcn"tgoctmu"cpf"eqpenwukqpu" The local volume averaging as founded by Anderson and Jackson, Slattery, Whitaker, was applied to derive mass conservation equations for multi-phase multi-component flows conditionally divided into three velocity fields. The heterogeneous porous-media formulation introduced by Gentry, Martin and Daly, commented by Hirt, and used by Sha, Chao and Soo, was then implanted into the formalism as a geometrical skeleton because of its practical usefulness. Beyond these concepts, inert components in each velocity field are included, and the concept of dynamic particle fragmentation and coalescence is introduced. Then subsequent time averaging was performed.
36
1 Mass conservation
The result of this derivation yields the following three local volume and time average equations applicable for each velocity field. Here, the equations are written in the scalar form for the most frequently used Cartesian and cylindrical coordinate systems to simplify their direct use by the reader for his particular application.
∂ 1 ∂ 1 ∂ ∂ (α l ρlγ v ) + κ ( r κ α l ρl ulγ r ) + κ (α l ρl vlγ θ ) + (α l ρl wlγ z ) ∂τ r ∂r r ∂θ ∂z = γv
lmax , w
∑ (μ m =1
ml
− μlm ) ,
(1.154)
∂ 1 ∂ ⎡ κ ∂ Cil ⎛ r α ρ u C − Dil* (α ρ C γ ) + ∂τ l l il v r κ ∂ r ⎢⎣ l l ⎜⎝ l il ∂r +
1 ∂ ⎡ 1 ∂ Cil ⎛ α l ρl ⎜ vl Cil − Dil* κ κ ⎢ r ∂θ ⎣ r ∂θ ⎝
= γv
lmax , w
∑ (μ m =1
iml
⎞ ⎤ ⎟γ r ⎥ ⎠ ⎦
∂ Cil ⎞ ⎤ ∂ ⎡ ⎛ α l ρl ⎜ wl Cil − Dil* ⎟γθ ⎥ + ⎢ ∂z ⎠ ⎦ ∂z⎣ ⎝
− μilm ) for α l ≥ 0 ,
⎞ ⎤ ⎟γ z ⎥ ⎠ ⎦
(1.155)
ν tn ∂ n ⎞ ⎤ 1 ∂ ⎡⎛ ∂ 1 ∂ ⎡ ⎛ νt 1 ∂n ⎞ ⎤ ( nl γ v ) + κ ⎢ r κ ⎜ ul nl − l tn l ⎟ γ r ⎥ + κ ⎢⎜ vl nl − l t κ l ⎟ γ θ ⎥ Scl r ∂θ ⎠ ⎦ ∂τ r ∂ r ⎢⎣ ⎝ Scl ∂ r ⎠ ⎥⎦ r ∂θ ⎣⎝ +
ν lt ∂ nl ⎞ ⎤ ∂ ⎡⎛ ⎢⎜ wl nl − t ⎟ γ z ⎥ = γ v ( nl ,kin − nl ,coal + nl ,sp ) ∂ z ⎣⎝ Scl ∂ z ⎠ ⎦
⎧1 ⎫ 1 ≡ γ v nl ⎨ nl ,kin + fl ,sp − ⎡⎣( f ds,col + f dno,col ) Pdno,coal + f do,col Pdo,coal ⎤⎦ − f l ,spectrum _ cut ⎬ n 2 ⎩ l ⎭ for α l ≥ 0 . (1.156)
For flow in a pipe the entrainment and deposition terms have to be taken into account. The link between kinetic generation of particles and the mass source terms is specified. The link between mass sources and the change in the bubble/droplet size due to evaporation or condensation was presented for local volume and timeaveraged source terms. The concept of mono-dispersity is discussed and a method is proposed for computation of the disappearance of particles due to evaporation and condensation. The following conclusion can be drawn from this derivation: 1. For numerical integration, the size of discretization for the above local volume and time average equations is allowed to be smaller than, equal to, or larger than the characteristic length scale Dl of the velocity field, unless other mathematical constraints associated with the numerical method used or concept used for the computation of the fragmentation and coalescence sources are imposed. The
References
37
characteristic length scale Dl is understood to be: Dl = 2 / κ l , for continuum, 2.
3. 4. 5.
and Dl = particle size, for disperse field. The size of the spatial discretization for the resulting equations should not be an order of magnitude larger than the characteristic length scale of the fragmentation. This ensures that important local events will not be obscured as a result of averaging of the properties in the finite volume. For example, insufficiently fine spatial discretization on modeling of melt-water interaction may lead to non-prediction of steam explosions for situations where explosions have been observed in experiments because of averaging over large volume having regions with intensive fragmentation and others without such fragmentation. The use of the concept of mono-dispersity is associated with the effect of disappearance of part of the particles during a given time step in the case of coincident mass transfer that reduces the local volume fraction of the phase. Kinetic particle source terms are directly related to mass source terms in the mass conservation equation. Non-kinetic particle source terms such as fragmentation or coalescence influence average particle size and, with the size, the amount of evaporation or condensation. This means that the non-kinetic particle source terms indirectly influence the mass of the field. These terms introduce important time delays compared to the approach where the quasi-static particle size under local conditions is used. This is of major significance in the analysis of severe accidents such as steam explosion or processes associated with condensation shocks.
Tghgtgpegu" Anderson TB, Jackson R (1967) A fluid mechanical description of fluidized beds, Ind. Eng. Fundam., vol 6 p 527 Batchelor GK (1970) The stress system in a suspension of force-free particles, J. Fluid Mech., vol 42 pp 545-570 Brackbill JU, Kothe DB, Zemach C (1992) A continuum method for modeling surface tension, J. Comput. Phys., vol 100 pp 335-354 Delhaye JM, Giot M, Reithmuller ML (1981) Thermohydraulics of two-phase systems for industrial design and nuclear engineering, Hemisphere Publishing Corporation, New York, McGraw Hill Book Company, New York Deich ME, Philipoff G A (1981) Gas dynamics of two phase flows. (In Russian) Energoisdat, Moscow Faeth GM (1995) Spray combustion: A Review, Proc. of The 2nd International Conference on Multiphase Flow ‘95 Kyoto, April 3-7, 1995, Kyoto, Japan Fick A (1855) Über Diffusion. Ann. der Physik, 94 p 59 Gentry RA, Martin RE and Daly BJ (1966) An Eulerian differencing method for unsteady compressible flow problems, J. Comp. Physics, vol 1 p 87 Gray WG, Lee PCY (1977) On the theorems for local volume averaging of multi-phase system, Int. J. Multi-Phase Flow, vol 3 pp 222-340
38
1 Mass conservation
Grigorieva VA and Zorina VM (eds) (1988) Handbook of thermal engineering, thermal engineering experiment, in Russian, 2d Edition, Moskva, Atomisdat, vol 2, Griffith L (1943) A theory of the size distribution of particles in a comminuted system, Canadian J. Research, vol 21 A no. 6 pp 57-64 Hetstrony G (1982) Handbook of multi phase systems. Hemisphere Publ. Corp., Washington et al., McGraw-Hill Book Company, New York ... Hirt CW, Nichols BD (1981) Volume of fluid (VOF) method for dynamics of free boundaries, J. Comput. Phys., vol. 39 pp 201 – 225 Hirt CW (1993) Volume-fraction techniques: powerful tools for wind engineering, Journal of Wind Engineering and Industrial Aerodynmics, vol 46&47 pp 327-338 Ishii M (1975) Thermo-fluid dynamic theory of two-phase flow, Eyrolles, Paris Kataoka I, Ishii M, Mishima K (June 1983) Transactions of the ASME, vol 105 pp 230-238 Kocamustafaogulari G, Ishii M (1983) Interfacial area and nucleation site density in boiling systems, Int. J. Heat Mass transfer, vol 26 no 9 pp 1377-1387 Kolev NI (March 1985a) Transiente Drephasen Dreikomponenten Strömung, Teil 1: Formulierung des Differentialgleichungssystems, KfK Report 3910 Kolev NI (August 1985b) Transiente Dreiphasen Dreikomponenten Stroemung, Teil 2: Eindimensionales Schlupfmodell Vergleich Theorie-Experiment, KfK Report 3926 Kolev NI (1986a)Transiente Dreiphasen Dreikomponenten Strömung, Teil 3: 3D-DreifluidDiffusionsmodell, KfK Report 4080 Kolev N I (1986b) Transient three-dimensional three-phase three-component nonequilibrium flow in porous bodies described by three-velocity fields, Kernenergie 29 pp 383-392 Kolev NI (August 1987) A three field-diffusion model of three-phase, three-component flow for the tansient 3D-computer code IVA2/001. Nuclear Technology, vol 78 pp 95-131 Kolev NI (1990) Derivatives for the state equations of multi-component mixtures for universal multi-component flow models, Nuclear Science and Engineering, vol 108 pp 74-87 Kolev NI (Sept. 1991a) A three-field model of transient 3D multi-phase, three-component flow for the computer code IVA3, Part 1: Theoretical basics: Conservation and state equations, numerics, KfK Report 4948 Kolev NI (1991b) IVA3: A transient 3D three-phase, three-component flow analyzer, Proc. of the Int. Top. Meeting on Safety of Thermal Reactors, Portland, Oregon, July 21-25, 1991, pp 171-180. Presented at the 7th Meeting of the IAHR Working Group on Advanced Nuclear Reactor Thermal-Hydraulics, Kernforschungszentrum Karlsruhe, August 27 to 29 1991 Kolev NI (1993a) Fragmentation and coalescence dynamics in multi-phase flows, Experimental Thermal and Fluid Science, vol 6 pp 211-251 Kolev NI (1993b) The code IVA3 for modelling of transient three-phase flows in complicated 3D geometry, Kerntechnik, vol 58 no 3 pp 147-156 Kolev NI (1993c) IVA3 NW: Computer code for modeling of transient three phase flow in complicated 3D geometry connected with industrial networks, Proc. of the Sixth Int. Top. Meeting on Nuclear Reactor Thermmal Hydraulics, Oct. 5-8, 1993, Grenoble, France Kolev NI (1993d) Berechnung der Fluiddynamischen Vorgänge bei einem Sperrwasserkühlerrohrbruch, Projekt KKW Emsland, Siemens KWU Report R232/93/0002 Kolev NI (1993e) IVA3-NW A three phase flow network analyzer. Input description, Siemens KWU Report R232/93/E0041 Kolev NI (1993f) IVA3-NW Components: Relief Valves, Pumps, Heat Structures, Siemens KWU Report R232/93/E0050
References
39
Kolev NI (1994a) The influence of the mutual bubble interaction on the bubble departure diameter, Experimental Thermal and Fluid Science, 8 pp 167-174 Kolev NI (1994b) The code IVA4: Modeling of mass conservation in multi-phase multicomponent flows in heterogeneous porous media, Kerntechnik, vol 59 no 4-5 pp 226-237 Kolev NI (1996) Three fluid modeling with dynamic fragmentation and coalescence fiction or daily practice? 7th FARO Experts Group Meeting Ispra, October 15-16, 1996; Proceedings of OECD/CSNI Workshop on Transient thermal-hydraulic and neutronic codes requirements, Annapolis, MD, U.S.A., 5th-8th November 1996; 4th World Conference on Experimental Heat Transfer, Fluid Mechanics and Thermodynamics, ExHFT 4, Brussels, June 2-6, 1997; ASME Fluids Engineering Conference & Exhibition, The Hyatt Regency Vancouver, Vancouver, British Columbia, CANADA June 22-26, 1997, Invited Paper; Proceedings of 1997 International Seminar on Vapor Explosions and Explosive Eruptions (AMIGO-IMI), May 22-24, Aoba Kinen Kaikan of Tohoku University, Sendai-City, Japan Kolev NI (1998) Can melt-water interaction jeopardize the containment integrity of the EPR? Part 3: Fragmentation and coalescence dynamics in multi-phase flows, KWU NA-T/1998/E083a, Project EPR Kolev NI (1999) Verification of IVA5 computer code for melt-water interaction analysis, Part 1: Single phase flow, Part 2: Two-phase flow, three-phase flow with cold and hot solid spheres, Part 3: Three-phase flow with dynamic fragmentation and coalescence, Part 4: Three-phase flow with dynamic fragmentation and coalescence – alumna experiments, CD Proceedings of the Ninth International Topical Meeting on Nuclear Reactor Thermal Hydraulics (NURETH-9), San Francisco, California, October 3-8,1999, Log. Nr. 315 Kolomentzev AI, Dushkin AL (1985) Vlianie teplovoj i dinamiceskoj neravnovesnosty faz na pokasatel adiabatj v dwuchfasnijh sredach, TE 8 pp 53-55 Mugele R A, Evans H D (151) Droplet size distribution in sprays, Ing. Eng. Chem., vol.43 pp 1317-1324 Nukiama S, Tanasawa Y (1931) Trans. Soc. Mech. Engrs. (Japan), vol 4 no 14 p.86 Pilch M, Erdman CA, Reynolds AB (August 1981) Acceleration induced fragmentation of liquid drops, Charlottesville, VA: Department of Nucl. Eng., University of Virginia, NUREG/CR-2247 Reid RC, Prausnitz JM, Sherwood TK (1982) The properties of gases and liquids, Third Edition, McGraw-Hill Book Company, New York Reynolds O (May 1894) On the dynamical theory of incompressible viscous fluids and the determination of the criterion, Cambridge Phil. Trans., pp 123-164 Riznic J R, Ishii M (1989) Bubble number density and vapor generation in flushing flow, Int. J. Heat Mass Transfer, vol 32 no 10 pp 1821-1833 Rosin P, Rammler E, (1933) Laws governing the fineness of powdered coal, J. Inst. Fuel, vol 7 pp 29-36 Sauter J (1929) NACA Rept. TM-518 Serizawa A, Kataoka I and Michiyoshi III (1975) Turbulence structure of air-water bubbly flow – I. Transport properties, Int. J. Multiphase Flow, Vol 2 pp 247-259 Sha T, Chao BT, Soo SL (1984) Porous-media formulation for multi-phase flow with heat transfer, Nuclear Engineering and Design, vol 82 pp 93-106 Slattery JC (1967) Flow of viscoelastic fluids through porous media, AIChE Journal, vol 13 p 1066 Slattery JC (1990) Interfacial transport phenomena, Springer, Berlin Heidelberg New York Teletov SG (1958) On the problem of fluid dynamics of two-phase mixtures, I. Hydrodynamic and energy equations, Bulletin of the Moscow University, No 2, p 15
40
1 Mass conservation
Wallis GB (1969) One-dimensional two-phase flow, McGraw-Hil, New York 1969 Whitaker S (1967) Diffusion and dispersion in porous media, AIChE Journal, vol 13 p 420 Whitaker S (1969) Advances in theory of fluid motion in porous media, Ind. Eng. Chem., vol 61 no 12 pp 14-28 Whitaker S (1977) Experimental principles of heat transfer, Pergamon Press Inc., New York, 1977 Whitaker S (1985) A simple geometrical derivation of the spatial averaging theorem, Chemical Engineering Education, Chemical Engineering Education, pp 18-21 pp 50-52 Zemansky MW (1968) Heat and thermodynamics, McGraw-Hill Book Company, Fifth Edition
4"Oqogpvwou"eqpugtxcvkqp"
The time rate for change in momentum of a body equals the net force exerted on it. Isaac Newton, Philosophiae naturalis principia, 1687
403"Kpvtqfwevkqp" As in Chapter 1, from the large number of formulations of the conservation equations for multi-phase flows the local volume averaging as founded by Anderson and Jackson, Slattery, and Whitaker was selected to derive rigorously the momentum equations for multi-phase flows conditionally divided into three velocity fields. The heterogeneous porous-media formulation introduced by Gentry et al., commented by Hirt, and used by Sha, Chao and Soo, is then implanted into the formalism as a geometrical skeleton. Beyond these concepts, I perform subsequent time averaging. This yields a working form which is applicable to a large variety of problems. All interfacial integrals are suitably transformed in order to enable practical application. Some minor simplifications are introduced in the general equation finally obtained, and working equations for each of the three velocity fields are recommended for general use in multiphase fluid dynamic analysis. This Chapter is an improved and extended version of the work published in Kolev (1994b). The strategy we follow is: We first apply the momentum equations for each of the velocity fields excluding the interfaces by replacing their actions by forces. Then, we write a force balance at the interfaces considering them as immaterial and therefore inertialess. This interfacial force balance links the momentum equations valid for the both sides of the interface.
42
2 Momentums conservation
404"Nqecn"xqnwog/cxgtcigf"oqogpvwo"gswcvkqpu" 40403"Ukping/rjcug"oqogpvwo"gswcvkqpu" The time rate for change in momentum of a body relative to an inertial frame of reference equals the net force exerted on it (Newton). Applied to a single continuum this principle results in Euler’s first law of continuum mechanics Truesdell (1968). This principle applied to each velocity field within the control volume except the interface yields the well-known local instantaneous momentum equation
∂ ( ρ Vτ ) + ∇ ⋅ ( ρl Vlτ Vlτ − Tlτ ) + ∇plτ + ρl g = 0, ∂τ l l
(2.1)
which is valid only inside the velocity field l excluding the interface. Here the positive velocity direction gives the negative force direction - a commonly used definition. The total stress tensor is split into plτ I and Tlτ . plτ is the static pressure inside field l, I is the unit matrix, and Tlτ is the shear stress tensor. Vlτ Vlτ is the dyadic product of two vectors Vlτ and Vlτ - see Appendix 1. It is a second order tensor. g is the vector of the gravitational acceleration. Equation (2.1) is the generally accepted momentums balance for single-phase for velocities much smaller then the velocity of light. 40404"Kpvgthceg"hqteg"dcncpeg"*oqogpvwo"lwor"eqpfkvkqp+" Next we abstract a volume around the interface having thickness ε converging to zero. Decoupling mechanically the control volume from the adjacent fields requires replacing the action of the forces on the volume by equivalent forces. In this case the Cauchy’s lemma holds: The stress vectors acting upon opposite sides of the same surface at a given point are equal in magnitude and opposite in direction, Truesdell (1968) p. 32. Characterization of the interface, surface tension: Consider two fluids with different densities, fluid m and fluid l as presented on Fig. 2.1.
2.2 Local volume-averaged momentum equations
43
f lm,nσ Fluid m
nl
δA
S
ds
rs
C
z
tl
y f lm,t σ
Fluid l
x
Fig. 2.1 Surface force - geometry definitions
Fluid m is a gas and fluid l is a liquid, ρl > ρ m . A three-dimensional surface S, de-
scribed by the position vector rS ( x, y , z ) , separates both fluids. We call such fluids imiscible. The unit normal vector of the liquid interface nl is pointing outside of the liquid l and is an important local characteristic of this surface. It is defined from the spatial distribution of the volume fraction of the field l as follows nl = −
∇ (α lγ )
∇ (α lγ )
.
(2.2)
Due to different molecular attraction forces at the two sides of the surface a resulting attraction force with special properties arises at the surface. The force exists only at the surface and acts at the denser fluid l. This force is called surface force. The surface force per unit mixture volume is denoted with flmσ . The subscript l indicates that the force acts at the field l and the subscript mσ indicates that the surface is an interface with filed m. We extract from the surface S an infinitesimal part δ A around the point rS ( x, y, z ) so that δA = nlδ A.
(2.3)
The closed curve C contains δ A. The closed curve C is a oriented counter clock wise. Consider the infinitesimal directed line element ds called arc length vector. The unit tangent vector to the surface S at C that is perpendicular to ds is t. In this case the relation holds tds = ds × n l .
(2.4)
The surface force exerted on the surface δ A by the surface outside of δ A across the directed line element ds is equal to σ lm t ds. Here σ lm is a material property being a force tangential to S per unit length called surface tension. It may vary
44
2 Momentums conservation
with the surface properties like temperature, concentration of the impurities of the liquid etc. The net surface force on the element δ A is then obtained by summing all forces σ lm tds exerted on each element of arc length ds, ∫ σ lm tds . Using C
Eq. (2.4) gives
∫σ
C
lm
tds = ∫ σ lm ds × nl .
(2.5)
C
The Stokes theorem, see Thomas et al. (1998) for derivation, allows one to transfer the integral over a closed curve to an integral over the surface closed by this curve
∫ ds × n σ l
C
lm
= ∫∫ ( nl × ∇ ) × nl σ lm dA.
(2.6)
S
We see that for the infinitesimal surface δ A the surface force per unit interface is
( nl × ∇ ) × nlσ lm = σ lm ⎡⎣( nl × ∇ ) × nl ⎤⎦ + ( nl × ∇σ lm ) × nl .
(2.7)
This force can be split into a normal and a tangential component by splitting the gradient into a sum of normal and tangential components ∇ = ∇ n + ∇t ,
(2.8)
where ∇ n = ( ∇ ⋅ nl ) nl
(2.9)
and ∇t = ∇ − ∇ n .
(2.10)
Using this splitting and after some mathematical manipulation Brackbill et al. (1992) simplified Eq. (2.7) and obtained finally the very important result,
( n l × ∇ ) × n lσ
= −n lσ lm ( ∇ ⋅ n l ) + ∇ tσ lm = n lσ lmκ l + ∇ tσ lm ,
(2.11)
where ⎡ ∇ (α lγ ) ⎤ ⎥ ⎣⎢ ∇ (α lγ ) ⎦⎥
κ l = − ( ∇ ⋅ nl ) = ∇ ⋅ ⎢
(2.12)
is the curvature of the interface defined only by the gradient of the interface unit vector. Equation (2.12) is used in Eq. (5.4), p. 23 in Drazin and Reid (1981). Note that the mathematical definition of curvature is the sum of the two principal curvatures which are magnitudes of two vectors. This sum is always positive. The expression resulting from Eq. (2.12) defines curvature with sign, which means - with its orientation. The curvature is positive if the center of the curvature is in the fluid m. In other words the positive curvature κ l is oriented along the normal vector nl .
2.2 Local volume-averaged momentum equations
45
Let us as an example estimate the curvature of a liquid layer in stratified flow between two horizontal planes with a gap equal to H. The interface is described by z the curve z* = = α 2 ( x ) being in the plane y = const. We use as coordinates H ( x, y, z *) . The gradient of the liquid volume fraction is then
∇α 2 ( x ) =
∂α 2 ( x ) ∂α ( x ) ∂α ( x ) ∂α ( x ) ∂z * i+ 2 k= 2 i+ k= 2 i + k. ∂x ∂z * ∂x ∂z * ∂x
The magnitude of the gradient is then 1/ 2
⎧⎪ ⎡ ∂α ( x ) ⎤ 2 ⎫⎪ ∇α 2 ( x ) = ⎨1 + ⎢ 2 ⎥ ⎬ . ⎪⎩ ⎣ ∂x ⎦ ⎪⎭
The normal vector is ⎛ ⎜ ∂α 2 ( x ) ⎜ ∇α 2 ( x ) ∂x n2 = − = −⎜ i+ 2 1/ 2 ⎜⎧ ∇α 2 ( x ) ⎫⎪ ⎡ ∂ α x ⎤ ( ) ⎪ 2 ⎜ ⎨1 + ⎜ ⎪ ⎢⎣ ∂x ⎥⎦ ⎬⎪ ⎭ ⎝⎩
⎞ ⎟ ⎟ 1 ⎟ k 1/ 2 ⎟ ⎧⎪ ⎡ ∂α ( x ) ⎤ 2 ⎫⎪ 2 ⎟ ⎨1 + ⎢ ⎥ ⎬ ⎟ ∂ x ⎣ ⎦ ⎩⎪ ⎭⎪ ⎠
The curvature in accordance with Eq. (2.12) is then ⎡ ∇ (α lγ ) ⎤ ⎥ ⎣⎢ ∇ (α lγ ) ⎦⎥
κ l = − ( ∇ ⋅ nl ) = ∇ ⋅ ⎢
⎛ ⎞ ⎜ ⎟ ∂α 2 ( x ) ⎜ ⎟ 1 ∂x ⎟ = ∇ ⋅⎜ i + k 2 1/ 2 2 1/ 2 ⎜⎧ ⎟ ⎫ ⎧ ⎫ ⎡ ∂ α x ⎤ ⎡ ∂ α x ⎤ ⎪ ⎪ ⎪ 2( ) 2 ( ) ⎜ ⎪⎨1 + ⎟ 1 + ⎨ ⎢ ⎥ ⎬ ⎜ ⎪ ⎢⎣ ∂x ⎥⎦ ⎬⎪ ∂ x ⎟ ⎦ ⎭⎪ ⎭ ⎩⎪ ⎣ ⎝⎩ ⎠ ∂α 2 ( x ) ∂ ∂ 1 ∂x = + . 1/ 2 2 ∂x ⎧ ⎡ ∂α x ⎤ ⎫ ∂z* ⎧ ⎡ ∂α x ⎤ 2 ⎫1/ 2 ( ) ( ) ⎪ ⎪ ⎪ ⎪ 2 2 ⎨1 + ⎢ ⎨1 + ⎢ ⎥ ⎬ ⎥ ⎬ ∂ x ∂ x ⎦ ⎪ ⎦ ⎭⎪ ⎩⎪ ⎣ ⎩⎪ ⎣ ⎭
46
2 Momentums conservation
Having in mind that ∂ 1 ∂z * ⎧ ⎡ ∂α x ⎤ 2 ⎫1/ 2 ⎪ ⎪ 2( ) ⎨1 + ⎢ ⎥ ⎬ ∂ x ⎦ ⎭⎪ ⎩⎪ ⎣
∂α 2 ( x ) ∂ 2α 2 ( x ) ∂x ∂x∂z * = 0 =− 3/ 2 ⎧⎪ ⎡ ∂α ( x ) ⎤ 2 ⎫⎪ 2 ⎨1 + ⎢ ⎥ ⎬ ∂x ⎦ ⎪ ⎩⎪ ⎣ ⎭
one finally obtains the well-known expression 3/ 2
⎧⎪ ⎡ ∂α ( x ) ⎤ 2 ⎫⎪ 2 ⎨1 + ⎢ ⎥ ⎬ . ∂ x ⎦ ⎪ ⎩⎪ ⎣ ⎭
∂ 2α 2 ( x ) κ2 = ∂x 2
The higher pressure is in the fluid medium on the concave side of the interface, since surface force in case of constant surface tension is a net normal force directed toward the center of curvature of the interface. Note, that the term σ lmκ l becomes important for 1/ κ l < 0.001m.
(2.13)
The term ∇ tσ lm can be expressed as ∇tσ lm = ( ∇t Tl )
∂σ lm i ∂σ + ∑ ( ∇t Cil ) lm . ∂ Tl ∂ Cil i =1 max
(2.14)
These terms describes the well-known Marangoni effect. The grid density for computational analysis can be judged by comparing the size of the control volume Δx with the curvature. Obvious, if and only if the volume is small enough, Δx κ l < 1 ,
(2.15)
the curvature can be resolved by the computational analysis. No velocity variation across the interface: Consider the case for which there is no velocity variation across the interface (e.g. stagnant fluids) - Fig. 2.2. The interface pressure plmσ ,τ is the normal force per unit surface acting on field l. This force acts inside the field l in the immediate vicinity of the interface and in the opposite direction on the interface control volume. It is different from the surface force exerted by the pressure of the neighboring velocity field, pmlσ ,τ . If there are no other forces except pressure and surface tension we have plmσ ,τ n l + pmlσ ,τ n m + ∇lσ lm + σ lmκ l nl = 0.
(2.16)
2.2 Local volume-averaged momentum equations
47
pmlσ ,τ −
nl
2σ lm nl Rl
m − field
Rl > 0
plmσ ,τ
σ − interface
l − field
σ lm
σ lm
Fig. 2.2 Interface force equilibrium without mass transfer and Marangoni effect (Laplace)
This equation is known as the Laplace equation. If in addition we have viscous forces acting on the two sides the momentum balance is
(T
mσ ,τ l
− plmσ ,τ I ) ⋅ n l + ( Tmlσ ,τ − pmlσ ,τ I ) ⋅ n m −∇lσ lm − σ lmκ l nl = 0 .
(2.17)
Velocity variation across the interface: If mass is transferred from one to the other fluid for whatever reason, the interface moves in space not only convectively but also controlled by the amount of mass transferred between the fields – Fig. 2.3. In τ this case the interface velocity Vlm is not equal to the neighboring field velocities. The mass flow rate
( ρ w)lm = ρl ( Vlτ
τ − Vlm ) ⋅ nl
enters the interface control volume and exerts the force ρl Vlτ ( Vlτ − Vlmτ ) ⋅ nl per unit surface on it. Note that this force has the same direction as the pressure force inside the field l. Similarly, we have a reactive force ρ m Vmτ ( Vmτ − Vlmτ ) ⋅ n m exerted per unit surface on the control volume by the leaving mass flow rate. Assuming that the control volume moves with the normal component of the interface velocity, Vlmτ ⋅ nl , we obtain the following force balance − ρl Vlτ ( Vlτ − Vlmτ ) ⋅ n l + ( Tlmσ ,τ − plmσ ,τ I ) ⋅ n l
− ρ m Vmτ ( Vmτ − Vmτ σ ) ⋅ n m + ( − pτmσ I + Tmτ σ ) ⋅ n m −∇lσ lm − σ lmκ l nl = 0 . (2.18)
48
2 Momentums conservation σ − interface
pmlσ ,τ m − field
(V
τ
l
)
τ − Vlm ⋅ nl
(
nl
ε→0
τ Vlm
Vlτ τ Vlm
nm mσ ,τ l
)
τ − Vmτ Vmτ − Vlm ⋅nm
Vmτ
p
l − field
Fig. 2.3 Definition of the interface characteristics
This is the general form of the interfacial momentum jump condition. It is convenient to rewrite the above equation by using the mass jump condition at the interface
ρl (Vlτ − Vlmτ ) ⋅ nl + ρ m (Vmτ − Vlmτ ) ⋅ n m = 0 ,
(2.19)
which is Eq. (1.42). The result is τ ⎡ ρl (Vlτ − Vlm ) ( Vmτ − Vlτ ) + ( pmlσ ,τ − plmσ ,τ ) I + Tlmσ ,τ − Tmlσ ,τ − σ lmκ l ⎤⎦ ⋅ nl − ∇lσ lm = 0 ⎣ (2.20)
or
( ρ w )lm ( Vmτ − Vlτ ) + ( pmlσ ,τ − plmσ ,τ ) nl + ⎡⎣Tlmσ ,τ
− Tmlσ ,τ − σ lmκ l ⎤⎦ ⋅ nl − ∇lσ lm = 0.
(2.21) The projection of this force to the normal direction is obtained by scalar multiplication of the above equation with the unit vector n l . The result is
( ρ w )lm ( Vmτ − Vlτ ) ⋅ nl + pmlσ ,τ − ( ∇lσ lm ) ⋅ n l = 0 .
− plmσ ,τ − σ lmκ l + ⎡⎣( Tlmσ ,τ − Tmlσ ,τ ) ⋅ nl ⎤⎦ ⋅ nl (2.22)
Using the mass conservation at the interface we finally have an important force balance normal to the interface ⎛ 1
1⎞ lσ ,τ mσ ,τ mσ ,τ − Tmlσ ,τ ) ⋅ n l ⎤⎦ ⋅ nl ⎟ + pm − pl − σ lmκ l + ⎡⎣( Tl ⎝ ρ m ρl ⎠ − ( ∇lσ lm ) ⋅ n l = 0 . (2.23)
( ρ w )lm ⎜ 2
−
2.2 Local volume-averaged momentum equations
49
Neglecting all the forces except those caused by pressure and interfacial mass transfer results in the surprising conclusion that during the mass transfer the pressure in the denser fluid is always larger than the pressure in the lighter fluid independently on the direction of the mass transfer - Delhaye (1981), p. 52, Eq. (2.64). For the limiting case of no interfacial mass transfer and dominance of the pressure difference the velocity of the interface can be expressed as a function of the pressure difference and the velocities in the bulk of the fields. pmlσ ,τ − plmσ ,τ . ρl (Vmn,τ − Vl n ,τ )
Vlmn ,τ = Vl n ,τ −
(2.24)
This velocity is called contact discontinuity velocity. Replacing the discontinuity velocity with Eq. (1.42) we obtain
(V
l
n ,τ
− Vmn ,τ ) = 2
ρl − ρ m mσ ,τ ( pl − pmlσ ,τ ). ρl ρ m
(2.25)
For the case ρl >> ρ m we have the expected result that the pressure difference equals the stagnation pressure at the side of the lighter medium plmσ ,τ − pmlσ ,τ = ρ m (Vl n,τ − Vmn,τ ) . 2
(2.26)
40405"Nqecn"xqnwog"cxgtcikpi"qh"vjg"ukping/rjcug"" oqogpvwo"gswcvkqp"
The aim here is to average Eq. (2.1) over the total control volume - see Fig. 2.4.
(1 − γ v)Vol
Vol
m n lσ lmax
Vol l ∑ l =1
w , structures
Vmτ
Flσ
l
Fle
Eq. (2.1) valid for field l only
Eq. (2.28) valid
θ
z
Vlτ r
for Vol
Local volume averaging
Fig. 2.4 Definition regions for single-phase instantaneous momentum balance and the local volume average balance
50
2 Momentums conservation
The mathematical tools used to derive local volume-averaged field conservation equations for the property being any scalar, vectorial, or tensorial function of time and location are once again Slattery - Whitaker's spatial averaging theorem, together with the Gauss - Ostrogradskii theorem, and the general transport equation (Leibnitz rule), see Anderson and Jackson (1967), Slattery (1967), Whitaker (1967, 1985, 1969), Gray and Lee (1977). Applying the local volume average to Eq. (2.1), the following is obtained
∂ ( ρ Vτ ) + ∇ ⋅ ( ρl Vlτ Vlτ ) − ∇ ⋅ ( Tlτ ) + ∇plτ + ρl g = 0, ∂τ l l
(2.27)
or using Eqs. (1.28), (1.32) and (1.28), Kolev (1994), 1 ∂ ρl Vlτ + ∇ ⋅ ρl Vlτ Vlτ +∇ ⋅ pτl I − Tlτ + ∂τ Vol 1 − Vol
τ τ ∫ ( − p I + T ) ⋅ n dF + gρ l
l
l
l
∫
ρl Vlτ ( Vlτ − Vlσ ,τ ) ⋅ nl dF
Flσ + Flw
=0
(2.28)
Flσ + Flw
The volume average momentum equation can be rewritten using the weighted average, see Eq. (1.19), Kolev (1994), Vlτ
le , ρ
Vlτ
le
le
= ρl Vlτ
l
/ ρl ,
(2.29)
and le
Vlτ
le
= ρl Vlτ Vlτ
l
ρl .
(2.30)
The result is
( +∇ (α γ
∂ α lγ v ρl ∂τ e l
1 − Vol
l
Vlτ le
plτ
) + ∇ ⋅ (α γ
) +α γ
l v
∫ ⎡⎣− p
wσ ,τ l
Flw
le
e l
l
ρl g −
ρl 1 Vol
l
Vlτ
le
∫ ⎡⎣− p
Vlτ
mσ ,τ l
Flσ
le
) − ∇ ⋅ (α γ e l
Tlτ
le
)
τ I + Tlmσ ,τ − ρl Vlτ ( Vlτ − Vlm )⎤⎦ ⋅ nl dF
I + Tlwσ ,τ − ρl Vlτ ( Vlτ − Vlwτ ) ⎤⎦ ⋅ nl dF = 0.
(2.31)
We assume that the weighted average of products can be replaced by the products of the average. This should be borne in mind when constructing a numerical algorithm for solving the final system and selecting the size of the finite volume so as to be not so large as to violate the validity of this assumption. Note the differences between Eq. (2.31) and the final result obtained by Ishii (1975), Eq. (3.16): (a) the directional permeability is used here instead of the volumetric porosity in the pressure gradient term, and
2.3 Rearrangement of the surface integrals
51
(b) as for the mass conservation equation in Kolev (1994a), the volumetric porosity is kept below the time differential since it can be a function of time in a number of interesting applications. For the case γ v = γ = 1 and one-dimensional flow, Eq. (2.31) reduces to Eq. (15) derived by Delhaye in Hetstrony (1982), p. 163. Equation (2.31), the rigorously derived local volume average momentum equation, is not amenable to direct use in computational models without further transformation. In order to facilitate its practical use (a) the integral expression has to be evaluated, and (b) the time averaging must be performed subsequently. These steps, which go beyond Sha et al. (1984), are performed in Sections 2.3 and 2.4.
405"Tgcttcpigogpv"qh"vjg"uwthceg"kpvgitcnu" The expression under the surface integral is replaced by its equivalent from the momentum jump condition, Eq. (2.18): −
1 Vol
=−
τ τ τ στ στ ∫ ⎡⎣ − ρ V ( V − V ) + T − p I ⎤⎦ ⋅ n dF l
l
l
lm
m , l
m , l
l
Flσ
{
}
1 ⎡ − ρ m Vmτ ( Vmτ − Vmτ σ ) + Tmlσ ,τ − pmlσ ,τ I + σ lmκ l ⎤ ⋅ nl + ∇l σ lm dF . (2.32) ⎦ Vol F∫lσ ⎣
Note that there are no surface force terms in Eq. (2.31). Equation (2.32) reflects the action of the surface forces and the action of the stresses caused by the surrounding field m on l. Note also that if the momentum equations are written for two neighboring fields the interface forces will appear in both equation with opposite sign only if the equations are applied to a common dividing surface. Otherwise, the exchange terms in question will be non-symmetric. The intrinsic surface-averaged field pressure at the entrances and exits of the control volume crossing the field m is pτm
me
. We call it bulk pressure inside the
velocity field m at this particular surface. The interfacial m-side pressure pmlσ ,τ can be expressed as the sum of the intrinsic averaged pressure, pτm
me
, which is not a
function of the position at the interface inside the control volume and can be taken outside of the integral sign, and a pressure difference Δpmlσ ,τ , which is a function of the position at the interface in the control volume pmlσ ,τ = pτm
me
+ Δpmlσ ,τ .
(2.33)
52
2 Momentums conservation
The same is performed for all other fields. Similarly, the surface pressure of the field structure interface is expressed as le
plwσ ,τ = plτ
+ Δplwσ ,τ .
(2.34)
The interfacial pressure differs from the bulk pressures of the corresponding fields mσ , and lw, Δpmlσ ,τ ≠ 0 and Δplwσ ,τ ≠ 0 . This occurs because obstacles to the continuous phase can cause local velocity decreases or increases at the interface velocity boundary layer, resulting in increased or decreased pressure relative to bulk pressure. In order to estimate the surface integrals the exact dependence of the pressure as a function of the position at the interface inside the control volume has to be elaborated for each idealized flow pattern and form of the structure. The same is valid for the viscous shear stresses Tmlσ ,τ and Tlwσ ,τ . Note, that in order to estimate the interfacial pressure integrals for practical use of the momentum equations the time averaging must be performed first to make it admissible to use real pressure distributions measured experimentally on bodies in turbulent flow. Substituting Eqs. (2.32) to (2.34) into Eq. (2.31) we obtain
( + ∇ (α γ
∂ αγ ρ ∂τ l v l
−
1 Vol 1 Vol
le
Vlτ le
plτ
e l
−
l
) + ∇ ⋅ (α γ e l
) +αγ l
∫ {⎡⎣ −Δp
lσ ,τ m
Flσ
∫ ⎡⎣ −Δp
wσ ,τ l
Flw
l
v
ρl g +
ρl
l
Vlτ
le
le
Vlτ
1 pτm Vol F∫lσ
me
) − ∇ ⋅ (α γ
nl dF +
e l
Tlτ
1 plτ Vol F∫lw
le
)
le
n l dF
}
τ I + Tmlσ ,τ − ρ m Vmτ ( Vmτ − Vlm ) + σ lmκ l ⎤⎦ ⋅ nl + ∇tσ lm dF
I + Tlwσ ,τ − ρl Vlτ ( Vlτ − Vlwτ ) ⎤⎦ ⋅ nl dF = 0 .
Keeping in mind that
me
pτm
and
plτ
le
(2.35)
are not functions of the interface posi-
tion inside the control volume, and making use of Eq. (1.32) (see also Kolev (1994)), the bulk pressure integrals can then be rewritten as follows: pτm
me
1 Vol
= − pτm
me
le
∫ n dF +
plτ
∇ (α leγ ) +
(p
l
Flσ
τ
l
1 nlw dF Vol F∫lw
le
− pτm
me
) Vol1 ∫ n dF. l
(2.36)
Flw
Rearranging the surface tension integrals: The surface force per unit volume of the mixture is in fact the local volume average of the surface tension force flmσ = −
1 ( nlσ lmκ + ∇tσ lm )dF . Vol ∫∫ Flσ
(2.37)
2.3 Rearrangement of the surface integrals
53
Note that the orientation of this force is defined with respect the coordinate system given in Fig. 2.1. Using Eq. (1.29) we have flmσ = −σ lmκ l
1 Vol
∫∫ n σ l
κ dF −
lm
Flσ
⎡ 1 = σ lmκ l ⎢∇ (α leγ ) + Vol ⎣⎢
1 ( ∇tσ lm )dF Vol ∫∫ Flσ
⎤
1 ∫ n dF ⎥⎥ − Vol ∫∫ ( ∇ σ )dF , l
Flw
⎦
t
(2.38)
lm
Flσ
with normal ⎡ ⎤ ⎤ ⎡ ∇ (α lγ ) ⎤ ⎡ 1 1 e flm,nσ = σ lmκ l ⎢∇ (α leγ ) + nl dF ⎥ = σ lm ∇ ⋅ ⎢ nl dF ⎥ ⎥ ⎢∇ (α l γ ) + ∫ ∫ Vol Flw Vol Flw ⎣⎢ ∇ (α lγ ) ⎦⎥ ⎣⎢ ⎣⎢ ⎦⎥ ⎦⎥
for α l > 0
(2.39)
and tangential flm,tσ = −
1 ( ∇tσ lm )dF ≈ −alm∇tσ lm Vol ∫∫ Flσ
(2.40)
surface force components per unit volume of the mixture, respectively. Here alm is the interfacial area density. The local volume averaged surface force is called some times in the literature continuum surface force or abbreviated CSF, see Brackbill et al. (1992). Note that if the surface tension is a constant in space there is no resulting tangential force component. At plane surfaces the curvature is zero and therefore there is no normal force acting at the liquid. If the liquid consists of clouds of spheres the local surface force creates only a difference in pressures inside and outside the sphere, but there is not a net force influencing the total movement either of a single droplet or of the cloud of the droplets in the space due to this force. We express this fact by multiplying the surface force by the function δ1 being one for continuum and zero for disperse field. Using Eqs. (2.36) and (2.38) the pressure and the surface tension terms can be rearranged as follows
(
∇ ⋅ α leγ plτ −
le
) + Vol1 ∫
pτm
Flσ
1 ( nlσ lmκ + ∇tσ lm ) dF Vol ∫∫ Flσ
me
n l dF +
1 Vol
∫
Flw
plτ
le
nl dF
54
2 Momentums conservation
(
le
= ∇ ⋅ α leγ plτ
)−
pτm
me
∇ (α leγ ) +
(p
τ le
l
me
− pτm
) Vol1 ∫ n dF l
Flw
⎡ ⎤ 1 1 +δ lσ lmκ l ⎢∇ (α leγ ) + nl dF ⎥ − δ l ( ∇tσ lm )dF ∫ ∫∫ Vol Vol ⎢⎣ ⎥⎦ Flw Flσ le
= α leγ ∇ plτ 1 − δl Vol
+
(p
τ le
l
− pτm
me
)
⎛ 1 + δ lσ lmκ l ⎜ ∇ (α leγ ) + ⎜ Vol ⎝
⎞
∫ n dF ⎟⎟ l
Flw
⎠
∫∫ ( ∇ σ ) dF . t
(2.41)
lm
Flσ
Note that
α leγ ∇⋅ pτl
le
stands in Eq. (2.41) for
(
∇ ⋅ α leγ plτ
le
)−
plτ
le
∇ ⋅ (α leγ ).
Rearranging the integrals defining interfacial momentum transfer due to mass transfer: Again using the mass jump condition at the interface, which is Eq. (1.42)
ρ m Vmτ ( Vmτ − Vlmτ ) = ρl Vmτ ( Vlτ − Vlmτ ) ,
(2.42)
the surface integral is rearranged as follows: −
1 τ ⎡ − ρ m Vmτ ( Vmτ − Vlm ⎤ ⋅ n l dF = 1 ) ∫ ⎣ ⎦ Vol Flσ Vol
τ τ τ ∫ ⎡⎣ ρ V ( V − V )⎤⎦ ⋅ n dF. l
m
l
lm
l
(2.43)
Flσ
For practical applications the mass source term is split into non-negative components as already explained in Chapter 1 (see also Kolev (1994a)). In addition, it is assumed that the mass emitted from a field has the velocity of the donor field. As a result, the local volume-averaged interfacial forces related to mass transfer through the interfaces can then be rewritten as follows: a) The components related to mass injection into, or suction from, the field through the field-structure interface are replaced by the “donor” hypothesis −
(
1 ⎡ ρl Vmτ ( Vlτ − Vlwτ ) ⎤ ⋅ nl dF = γ v μ τwl Vwτ ⎦ Vol F∫lw ⎣
we
τ − μlw Vlτ
le
).
(2.44)
2.4 Local volume average and time average
55
b) The components due to evaporation, condensation, entrainment and deposition are replaced by the “donor” hypothesis −
τ τ τ τ ∫ ⎡⎣ ρ V ( V − V σ )⎤⎦ ⋅ n dF = γ ∑ ( μ 3
1 Vol
l
l
l
l
l
v
Flσ
Vmτ
ml
m =1
me
le
τ − μlm Vlτ
). (2.45)
The sum of all interface mass transfer components is then
(
3,w
τ γ v ∑ μ ml Vmτ m =1
me
le
τ − μlm Vlτ
).
(2.46)
Thus, the final form obtained for the local volume average momentum equation is as follows,
(
∂ α lγ v ρl ∂τ + α leγ∇ plτ 1 − δl Vol
+
1 Vol
l
le
Vlτ
+
le
(p
τ
l
) + ∇ ⋅ (α γ e l
le
− pτm
1 ∫F (∇tσ lm )dF + Vol lσ
∫ ( Δp
wσ ,τ l
me
ρl
l
Vlτ
le
Vlτ
le
e l
)
⎛ 1 + δ lσ lmκ l ⎜ ∇ (α leγ ) + ⎜ Vol ⎝
∫ ( Δp
lσ ,τ m
lσ ,τ m
I−T
Flσ
3, w
(
m =1
Tlτ
le
) +α γ l
l
v
ρl g
⎞
∫ n dF ⎟⎟ l
⎠
Flw
) ⋅ n dF
τ I − Tlwσ ,τ ) ⋅ nl dF = γ v ∑ μ ml Vmτ
Flw
) − ∇ ⋅ (α γ
l
me
τ − μ lm Vlτ
le
),
(2.47)
and is independent of whether the field is structured or non-structured. Before we continue with the estimation of the remaining integrals, we will perform first a time averaging of Eq. (2.47).
406"Nqecn"xqnwog"cxgtcig"cpf"vkog"cxgtcig" The instantaneous surface-averaged velocity of the field l,
Vlτ
le
, can be ex-
pressed as the sum of the surface-averaged velocity which is subsequently timeaveraged, Vl = Vlτ
le
(2.48)
and a pulsation component Vl′ , Vlτ
le
= Vl + Vl ′,
(2.49)
56
2 Momentums conservation
as proposed by Reynolds. The fluctuation of the velocity is the predominant fluctuation component relative to, say, the fluctuation of α l or ρl . Introduction of Eq. (2.49) in the momentum conservation equation and the time averaging yields
∂ (αl ρl Vl γ v ) + ∇ ⋅ (αle ρl Vl Vl γ ) + ∇ ⋅ ⎡⎣α leγ ρl Vl′Vl′ − δ l Tl ⎤⎦ +α leγ∇pl + α lγ v ρl g ∂τ
(
⎛ 1 + ( pl − pm + δ lσ lmκ l ) ⎜ ∇ (α leγ ) + ⎜ Vol ⎝ +
1 Vol
∫ ( Δp
lσ m
I − Tmlσ ) ⋅ nl dF +
Flσ
3, w
= γ v ∑ ( μ kl Vk − μlk Vl k =1 k ≠l
1 Vol
)
⎞
∫ n dF ⎟⎟ − δ l
⎠
Flw
∫ ( Δp
wσ l
l
1 Vol
∫ ( ∇ σ )dF t
lm
Flσ
I − Tlwσ ) ⋅ nl dF
Flw
)
(2.50a)
see also Appendix 2.1. It is evident from Eq. (2.50a) that the products of the pulsation velocity components, called Reynolds stresses, act on the flow introducing additional macroscopic cohesion inside the velocity field. Equation (2.50a) is applied on the field l including the surface up to the m-side interface. For dispersed flows it is convenient to have also the momentum equation of the continuum in a primitive form without using the momentum jump condition. In this case the time average of Eq. (2.31) for field m after introducing Eq. (2.33) is
∂ (α m ρm Vmγ v ) + ∇ ⋅ (α me ρm Vm Vmγ ) + ∇ ⋅ ⎡⎣α me γ ρm Vm′ Vm′ − δ m Tm ⎤⎦ + α mγ v ρ m g ∂τ
(
+α me γ∇pm −
1 Vol
∫ ( Δp
lσ m
I − Tmlσ ) ⋅ nl dF +
Fmσ
1 Vol
)
∫ ( Δp
wσ m
I − Tmwσ ) ⋅ n m dF
Fmw
3, w
= γ v ∑ ( μ km Vk − μ mk Vm ) .
(2.50b)
k =1 k ≠m
Comparing Eq. (2.50a) with Eq. (2.50b) we realize that the term −
1 Vol
∫ ( Δp
lσ m
I − Tmlσ ) ⋅ nl dF
Fmσ
appears in the both equations with opposite sign. For practical computation we recommend the use of a couple of equations having common interface in order to easily control the momentum conservation at the selected common interface.
2.6 Viscous and Reynolds stresses
57
407"Fkurgtugf"rjcug"kp"ncokpct"eqpvkpwwo"/"rugwfq" vwtdwngpeg" It is known that even low-velocity potential flow over a family of spheres is associated with natural fluctuations of the continuum. The produced oscillations of the laminar continuum are called pseudo-turbulence by some authors. The averaged pressure over the dispersed particles surface is smaller then the volume averaged pressure. Therefore in flows with spatially changing concentration of the disperse phase an additional force acts towards the concentrations gradients. For bubbly flow Nigmatulin (1979) obtained the analytical expression
− Vc′Vc′ = ΔVcd
2
4 20
0
0
0
3 20
0 .
0
0
3 20
Van Wijngaarden (1998) used this expression multiplied by α d .
408"Xkueqwu"cpf"Reynolds uvtguugu" The solid-body rotation and translation of the fluid element does not cause any deformation and therefore no internal viscous stresses in the fluid. Only the deformation of the fluid element causes viscous stress resisting this deformation. For estimation of the relationship between deformation and viscous stress, the heuristic approach proposed by Helmholtz (1868) and Stokes (1845a and b) can be used for the continuous, intrinsic isotropic, non-structured field, see Schlichting (1959) p. 58. The background conditions behind this approach will now be recalled: (a) the field is a continuum, (b) small velocity changes are considered, (c) only the linear part of the Taylor series is taken into account, and (d) linear dependence between stresses and velocity deformations (Newtonian continuum). The mathematical notation for this hypothesis is 2 2 ⎡ T ⎤ ⎡ ⎤ Tη = η ⎢∇V + ( ∇V ) − ( ∇ ⋅ V ) I ⎥ = η ⎢ 2D − ( ∇ ⋅ V ) I ⎥ 3 3 ⎣ ⎦ ⎣ ⎦
58
2 Momentums conservation
⎛ ⎛ ∂u 1 ⎞ ∂v ∂u ∂w ∂u ⎞ − ∇⋅V⎟ + + ⎜ 2⎜ ⎟ ∂x ∂y ∂x ∂z ⎠ ⎜ ⎝ ∂x 3 ⎟ ⎜ ⎟ ⎛ ∂v 1 ⎞ ∂u ∂v ∂w ∂v ⎟. =η ⎜ + 2⎜ − ∇ ⋅V ⎟ + ∂y ∂x ∂y ∂z ⎜ ⎟ ⎝ ∂y 3 ⎠ ⎜ ⎟ ∂u ∂w ∂v ∂w ⎛ ∂w 1 ⎞⎟ ⎜ + + 2⎜ − ∇⋅V⎟⎟ ⎜ ∂z ∂x ∂z ∂y ⎝ ∂z 3 ⎠⎠ ⎝
(2.51)
where T is the second order tensor for the viscous momentum flux, ∇V is the dyadic product of the Nabla operator and the velocity vector (a second order tensor), T designates the transposed tensor. Note that Nabla operator of the velocity vector, ∇V = D + W
consists of a symmetric part D=
1⎡ T ∇V + ( ∇V ) ⎤ ⎦ 2⎣
called deformation rate and a skew part W=
1⎡ T ∇ V − ( ∇V ) ⎤ ⎣ ⎦ 2
called spin or vortices tensor. ∇⋅V =
∂u ∂v ∂w + + ∂x ∂y ∂z
is the divergence of the velocity vector. The term containing the divergence of the velocity vector is called by Stokes rate of cubic dilatation. The hypothesis says that the relation between viscous stresses and deformation rate on a control volume is linear and the proportionality factor is the dynamic viscosity η , that solid body translations and rotations do not contribute to the viscous forces, that the share stresses are symmetric, and that the relation between volumetric and share viscosity is so that the pressure always equals one third of the sum of the normal stresses. Each of this point has his ingenious argumentation in the Stokes paper. In the multiphase continuous fields, as long as they are fine resolved this stress tensor exists. I recommend to any one having serious intention to understand flows to study this paper. Alternatively see Schliching, (1959) p. 60, where it is explained that Eq. (2.51) contains the Stokes result for the relation of the bulk viscosity equal to -2/3 dynamic viscosity. From the mechanical equilibrium condition for all angular momenta around an axis for vanishing dimensions of the control volume, one obtains the symmetry of the components of the viscous stress tensor, see Schlichting (1959), p. 50 for Cartesian coordinates.
2.6 Viscous and Reynolds stresses
59
For practical use it is convenient to write the viscous stress tensor for Cartesian and cylindrical coordinates as follows ⎛ ⎡ ∂u 1 ⎤ ⎜ 2 ⎢ − (∇ ⋅ V )⎥ ∂ r 3 ⎦ ⎜ ⎣ ⎜ 1 ∂u v ∂v Tη = η ⎜ κ −κ κ + ∂r r ⎜ r ∂θ ⎜ ∂u ∂w ⎜⎜ + ∂z ∂r ⎝
1 ∂u v ∂v −κ κ + κ ∂θ ∂r r r v 1 ⎛ 1 ∂v ⎞ 2⎜ κ + κ κ − (∇ ⋅ V ) ⎟ 3 r ⎝ r ∂θ ⎠ ∂v 1 ∂w + ∂z r κ ∂θ
⎞ ⎟ ⎟ ∂v 1 ∂w ⎟ + ⎟, ∂z r κ ∂θ ⎟ ⎡ ∂w 1 ⎤⎟ 2⎢ − ( ∇ ⋅ V ) ⎥ ⎟⎟ ⎣ ∂z 3 ⎦⎠ ∂u ∂w + ∂z ∂r
where ∇⋅V =
( )
1 ∂ κ 1 ∂v ∂w r u + κ + . κ ∂r r r ∂θ ∂z
For cylindrical coordinates κ = 1. For Cartesian coordinates set κ = 0 and replace r , θ , z with x, y, z, respectively. Now let us step to the turbulence stress tensor − ρl Vl′Vl′ called Reynolds stress tensor. The search for a quantitative estimation of the Reynolds stresses for multiphase flows is in its initial stage. A possible step in the right direction, in analogy to single-phase turbulence, is the use of the Boussinesq hypothesis (1877) for the turbulent eddies viscosity inside the velocity field. Boussinesq introducing the idea of turbulent eddy viscosity inside the velocity field so 2 ⎡ ⎤ − ρl Vl′Vl′ = ηlt ⎢ 2Dl − ( ∇ ⋅ Vl ) I ⎥ 3 ⎣ ⎦ ⎛ ⎛ ∂u 1 ⎞ ∂v ∂u ∂w ∂u ⎞ − ∇⋅V⎟ + + ⎜ 2⎜ ⎟ ∂x ∂y ∂x ∂z ⎠ ⎜ ⎝ ∂x 3 ⎟ ⎜ ⎟ ⎛ ∂v 1 ⎞ ∂u ∂v ∂w ∂v t ⎟, = ηl ⎜ + 2⎜ − ∇ ⋅ V ⎟ + ∂y ∂x ∂y ∂z ⎜ ⎟ ⎝ ∂y 3 ⎠ ⎜ ⎟ ∂u ∂w ∂v ∂w ⎛ ∂w 1 ⎞⎟ ⎜ + + 2 − ∇ ⋅ V ⎜ ⎟⎟ ⎜ ∂z ∂x ∂z ∂y ⎝ ∂z 3 ⎠ ⎠l ⎝
(2.52)
that it has the same structure as the Stokes hypothesis. The corresponding Reynolds stresses are ⎛ ∂u 1 ⎞ − ρl ul′ul′ = τ l′, xx = ηlt 2 ⎜ l − ∇ ⋅ Vl ⎟ , ⎝ ∂x 3 ⎠ ⎛ ∂v 1 ⎞ − ρl vl′vl′ = τ l′, yy = ηlt 2 ⎜ l − ∇ ⋅ Vl ⎟ , ⎝ ∂y 3 ⎠
60
2 Momentums conservation
⎛ ∂w 1 ⎞ − ρl wl′wl′ = τ l′, zz = ηlt 2 ⎜ l − ∇ ⋅ Vl ⎟ , ⎝ ∂y 3 ⎠ ⎛ ∂v ∂u ⎞ − ρl ul′vl′ = τ l′, xy = ηlt ⎜ l + l ⎟ , ⎝ ∂x ∂y ⎠
⎛ ∂w ∂u ⎞ − ρl ul′wl′ = τ l′, xz = ηlt ⎜ l + l ⎟ , ∂z ⎠ ⎝ ∂x ⎛ ∂w ∂v ⎞ − ρl vl′wl′ = τ l′, yz = ηlt ⎜ l + l ⎟ . ∂z ⎠ ⎝ ∂y
The dynamic turbulent viscosity now is a flow property and remains to be estimated. Note that at a given point this is a single value for all directions. Strictly speaking this approach is valid for isotropic turbulence because there is a single eddy viscosity assumed to be valid for all directions. An alternative notation of the term ∇ ⋅ ⎡α le ρl Vl′Vl′ γ ⎤ is given here for iso⎣ ⎦ tropic turbulence for which
(
)
2 kl : 3 ⎛ ∂γ xα le ρl ul′ul′ ∂γ yα le ρl ul′vl′ ∂γ zα le ρl ul′wl′ ⎞ + + ⎜ ⎟ ∂x ∂y ∂z ⎜ ⎟ e ⎜ ⎟ e e ∂γ xα l ρl vl′ul′ ∂γ yα l ρl vl′vl′ ∂γ zα l ρl vl′wl′ ⎟ ∇ ⋅ ⎡⎣(α le ρl Vl′Vl′) γ ⎤⎦ = ⎜ + + ⎜ ⎟ ∂x ∂y ∂z ⎜ ⎟ e e e ⎜ ∂γ xα l ρl wl′ul′ ∂γ yα l ρl wl′vl′ ∂γ zα l ρl wl′wl′ ⎟ + + ⎜ ⎟ ∂x ∂y ∂z ⎝ ⎠ ul′ul′ = vl′v′ = wl′wl′ =
⎛ ∂γ yα le ρl ul′vl′ ∂γ zα le ρ l ul′wl′ ⎞ + ⎜ ⎟ ∂y ∂z ⎜ ⎟ ⎜ ∂γ α e ρ v ′u ′ ∂γ α e ρ v′w′ ⎟ 2 e = ⎜ x l l l l + z l l l l ⎟ + ∇ ( γα l ρl kl ) 3 ∂x ∂z ⎜ ⎟ e ⎜ ⎟ e ′ ′ ∂ γ α ρ w v ′ ′ ∂ γ α ρ w u ⎜ x l l l l + y l l l l⎟ ⎜ ⎟ ∂x ∂y ⎝ ⎠ 2 = −S l + ∇ ( γα le ρl kl ) , 3
where
(2.53)
(2.54)
2.6 Viscous and Reynolds stresses
⎛ ∂ ⎡ ∂u ⎞ ⎤ ∂ ⎡ ∂u ⎞ ⎤ ⎞ e t ⎛ ∂v e t ⎛ ∂w ⎜ ⎢γ yα l ρlν l ⎜ l + l ⎟ ⎥ + ⎢γ zα l ρlν l ⎜ l + l ⎟ ⎥ ⎟ ⎝ ∂x ∂z ⎠ ⎦ ⎟ ⎝ ∂x ∂y ⎠ ⎦ ∂z ⎣ ⎜ ∂y ⎣ ⎜ ⎟ ∂ul ⎞ ⎤ ∂ ⎡ ∂vl ⎞ ⎤ ⎟ ∂ ⎡ e t ⎛ ∂vl e t ⎛ ∂wl . S l = ⎜ γ α ρ ν + + γ α ρ ν + ⎜ ∂x ⎢ x l l l ⎜⎝ ∂x ∂y ⎟⎠ ⎥ ∂z ⎢ z l l l ⎜⎝ ∂y ∂z ⎟⎠ ⎥ ⎟ ⎣ ⎦ ⎣ ⎦ ⎜ ⎟ ⎜ ∂ ⎡ ∂ul ⎞ ⎤ ∂ ⎡ ∂vl ⎞ ⎤ ⎟ e t ⎛ ∂wl e t ⎛ ∂wl + ⎜⎜ ⎢γ xα l ρlν l ⎜ ⎟ ⎥ + ⎢γ yα l ρlν l ⎜ ∂y + ∂z ⎟ ⎥ ⎟⎟ ⎝ ∂x ∂z ⎠ ⎦ ∂y ⎣ ⎝ ⎠⎦ ⎠ ⎝ ∂x ⎣
61
(2.55)
2 ∇ ( γα le ρl kl ) is considered as dispersion force 3 and is directly computed from the turbulent kinetic energy delivered by the turbulence model. It is important to emphasize that, in spite of the fact that several processes in single-phase fluid dynamics can be successfully described by the HelmholtzStokes and by the Boussinesq hypotheses; these hypotheses have never been derived from experiments or proven by abstract arguments. This limitation of the hypotheses should be borne in mind when they are applied. While the heuristic approach proposed by Helmholtz and Stokes, Eq. (2.51), is valid only for the continuous part of each velocity field, the Boussinesq hypothesis is useful for continuous and disperse velocity fields. This behavior is again described here by introducing for each velocity field the multiplier δ l in Eq.
Here the diagonal symmetric term
(2.50), where δ l = 0 for dispersed field and δ l = 1 for continuous non- structured fields. For a single field, δ l = 1, the description of the viscous and Reynolds stresses reduces to the widely accepted expression. It is thinkable to define the turbulent pressure as pl′ = ρl
2 1 kl = ρl ( u ′u ′ + v ′v ′ + w′w′ ) , 3 3
(2.56)
and to consider the term −α leγ∇pl′ as absorbed from the pressure term α leγ ∇pl and the term − pl′∇ (α leγ ) =
1 1 pl′nl dF + pl′n l dF Vol F∫lσ Vol F∫lw
(2.57)
as included in the pressure differences between the bulk pressure and boundary layer pressure. This could means that pl′ could have no longer to appear in the notation. Until the correctness of this agglomeration of the terms is not strictly proven I do not recommend it. It is interesting to note that from the kinetic theory for two colliding particles plus their added mass the following is obtained
62
2 Momentums conservation
1 ⎡ ⎤ ∇ (α leγ pl′ ) ≈ ∇ ⎢α leγ ( ρl + clvm ρ m ) Vl′2 ⎥ . 3 ⎣ ⎦
(2.58)
409"Pqp/gswcn"dwnm"cpf"dqwpfct{"nc{gt"rtguuwtgu" 40903"Eqpvkpwqwu"kpvgthceg" 2.7.1.1 3D flows Examples for the existence of continuous interfaces are the stratified pool, see Fig. 2.5, and annular pipe flow. The treatment of the interface depends very much on the numerical method used. If the numerical method is able to resolve the interface itself and the two attached boundary layers (see for instance Hirt and Nichols (1981) for more information), the interface momentum jump condition is the only information which is needed to close the mathematical description of the interface. If this is not the case, special treatment of the processes at the interface is necessary. In this Section we discuss some possibilities. ΔVlmn
θ
z
ΔVlm
nl
m − field l − field
ΔVlmt r
Fig. 2.5 Continuous interface
Consider a liquid (l) – gas (m) flow without mass transfer. The computational cells are so large that the surface is at best represented by piecewise planes at which the surface tension is neglected. The compressibility of the gas is much larger then the compressibility of the liquid. In this case we can assume that there is almost no difference between the bulk and liquid side interface pressure Δplmσ = 0.
(2.59)
The pressure change across the gas side boundary layer is then approximated by the stagnation pressure
2.7 Non-equal bulk and boundary layer pressures
Δpmlσ = ρ m (Vl n − Vmn ) = ρ m ΔVlmn sign (Vl n − Vmn ) , 2
2
63
(2.60)
- see Fig. 2.6. The normal velocity difference required in the above expression can be obtained by splitting the relative velocity vector at the interface ΔVlm into a component which is parallel to nl ⎛ ΔV ⋅ n ⎞ ΔVlmn = projnl ΔVlm = ⎜ lm l ⎟ nl = ( ΔVlm ⋅ nl ) nl , ⎝ nl ⋅ nl ⎠
(2.61)
with a magnitude ΔVlmn = ΔVlm ⋅ nl = ⎡⎣ nlx ( ul − um ) ⎤⎦ + ⎡⎣ nly ( vl − vm ) ⎤⎦ + ⎡⎣ nlz ( wl − wm ) ⎤⎦ 2
2
2
(2.62)
and a component orthogonal to n l , ΔVlmt = ΔVlm − projel ΔVlm = ΔVlm − ( ΔVlm ⋅ n l ) n l = ⎡⎣ Δulm − nlx ( ΔVlm ⋅ nl )⎤⎦ i + ⎡⎣ Δvlm − nly ( ΔVlm ⋅ n l ) ⎤⎦ j + ⎡⎣ Δwlm − nlz ( ΔVlm ⋅ n l ) ⎤⎦ k (2.63) m − field
l − field
Δpmlσ
Vl Vm
Fig. 2.6 Stagnation pressure in stratified flow y
γ v ρ lg
dy dy dα l = γ v ρ lg dz dα l dz
z
Fig. 2.7 Geodesic pressure force
In a similar way, the stagnation pressure difference at the field-structure interface can be estimated,
64
2 Momentums conservation
Δplwσ = ρl (Vl n − Vwn ) sign (Vl n − Vwn ) . 2
(2.64)
For the case of Δpmlσ ≈ const within Vol the following can be written 1 Vol
∫ Δp
lσ m
1 Vol
nl dF = Δpmlσ
Flσ
∫ n dF l
Flσ
⎡ ⎤ 1 = −Δpmlσ ⎢∇ (α leγ ) + n l dF ⎥ . ∫ Vol Flw ⎢⎣ ⎥⎦ (2.65)
For Vln > Vmn and decreasing α leγ in space, this force resists the field l. If there is no difference in the average normal velocities at the interface the above term is zero. If the tangential average velocity difference differs from zero, there is a tangential viscous shear force −
1 Vol
∫ T σ ⋅ n σ dF = c m
l
ml
ΔVlmt ΔVlmt .
(2.66)
Flσ
Here cml has to be computed using empirical correlation in case of the large scale of the cells not resolving the details of the boundary layer. For a wavy surface, the interface share coefficient should be increased by a component for form drag caused by the non-uniform pressure distribution, which results in additional tangential force. Similarly, the wall viscous shear stress is
−
1 Tlw ⋅ n lw dF = cwl Vlt Vlt , Vol F∫lw
(2.67)
where like in the previous case, cwl has to be computed using empirical correlation. Note that in case of stratified rectangular duct flow in the z direction - see Fig. 2.7, the change in the liquid thickness causes a lateral geodesic pressure force
γ v ρl g
dy dy dα l . = γ v ρl g dz dα l dz
(2.68)
Here y is the distance between the bottom of the duct and the center of mass of the liquid. In three-dimensional models this force automatically arises due to differences in the local bulk pressure having the geodesic pressure as a component. This force should be taken into account in one-dimensional models. If this force is neglected, the one-dimensional model will not be able to predict water flow in a horizontal pipe with negligible gas-induced shear. In the next section we consider this problem in more detail.
2.7 Non-equal bulk and boundary layer pressures
65
2.7.1.2 Stratified flow in horizontal or inclined rectangular channels
Geometrical characteristics: Stratified flow may exists in regions with such relative velocities between the liquid and the gas which does not cause instabilities leading to slugging. Some important geometrical characteristics are specified here – compare with Fig. 2.8. The perimeter of the pipe is then Per1w = 2 ( a + H ) ,
(2.69)
and the wetted perimeters for the gas and the liquid parts are Per1w = a + 2 ( H − δ 2 F ) = a + 2α1 H ,
(2.70)
Per2 w = a + 2δ1F = a + 2α 2 H .
(2.71)
1 H 2
δ 2F
α2
a
Fig. 2.8 Definition of the geometrical characteristics of the stratified flow
The gas-liquid interface median is then a, and the liquid level
δ 2F = α2 H .
(2.72)
The hydraulic diameters for the gas and the liquid for computation of the pressure drop due to friction with the wall are therefore Dh1 = 4α1 F / Per1w =
4α1aH , a + 2 H α1
Dh 2 = 4α 2 F / Per2 w =
4α 2 aH , a + 2 Hα 2
(2.73) (2.74)
and the corresponding Reynolds numbers Re1 =
α1 ρ1w1 4α1aH , η1 a + 2 Hα1
(2.75)
66
2 Momentums conservation
Re2 =
α 2 ρ 2 w2 4α 2 aH . η2 a + 2 Hα 2
(2.76)
Here F is the channel cross section and Per1w and Per2w are the perimeters wet by gas and film, respectively. If one considers the core of the flow the hydraulic diameter for computation of the friction pressure loss component gas-liquid is then Dh12 = 4α1 F / ( Per1w + a ) =
2α1aH a + H α1
(2.77)
and the corresponding Reynolds number Re1 =
α1 ρ1 w1 − w2 2α1aH . η1 a + H α1
(2.78)
The gas-wall, liquid-wall and gas-liquid interfacial area densities are a1w =
Per1w a + 2α1 H = , F aH
(2.79)
a2 w =
Per2 w a + 2α 2 H , = F aH
(2.80)
a12 =
a 1 = . F H
(2.81)
For the estimation of flow pattern transition criterion the following expression is some times required dα 2 1 = . dδ 2 F H
(2.82)
Using the geometric characteristics and the Reynolds numbers the interfacial interaction coefficients can be computing by means of empirical correlations as discussed in Volume II of this monograph.
Gravitational (hydrostatic) pressure variation across the flow cross section of horizontal pipe: In stratified flow the gravitation is a dominant force. The cross section averaged gas gravity pressure difference with respect to the interface is Δp12σ = ρ1 gα1
H . 2
(2.83)
The cross section averaged liquid gravity pressure difference with respect to the interface is Δp12σ = − ρ 2 gα 2
H . 2
(2.84)
2.7 Non-equal bulk and boundary layer pressures
67
The cross section averaged field pressures in terms of the interfacial pressure are then p1 = p12σ − Δp12σ = p12σ − ρ1 gα1
H , 2
(2.85)
p2 = p12σ − Δp12σ = p12σ + ρ 2 gα 2
H , 2
(2.86)
recalling the definition Eqs. (2.33) and (2.34). The averaged pressure in the cross section can be expressed as the cross section weighted averaged pressures inside the fields p = α1 p1 + α 2 p2 = α1 p12σ + α 2 p12σ + g
H 2 (α 2 ρ2 − α12 ρ1 ) 2
(2.87)
Neglecting the surface tension we obtain p ≈ pσ + g
H 2 (α 2 ρ2 − α12 ρ1 ), 2
(2.88)
pσ ≈ p − g
H 2 (α 2 ρ 2 − α12 ρ1 ). 2
(2.89)
or
The averaged pressure in the gas and in the liquid phase can be then expressed as a function of the system averaged pressure and geometrical characteristics by replacing pσ in Eqs. (2.85) and (2.86): p1 = p −
H gα 2 ρ 2
for α1 > 0
(2.90)
p2 = p +
H gα1 ρ 2
for α 2 > 0
(2.91)
where ρ = α1 ρ1 + α 2 ρ 2 is the homogeneous mixture density. The check α1 p1 + α 2 p2 = p proves the correctness of the computation. The difference between both averaged pressures is then Δp21 = p2 − p1 =
H g ρ for α1 > 0 and α 2 > 0 2
(2.92)
Delhaye (1981), p. 89, therefore p1 = p − α 2 Δp21,
(2.93)
p2 = p + α1Δp21 ,
(2.94)
and consequently
68
2 Momentums conservation
∂p1 ∂p ∂α ∂Δp21 = − Δp21 2 − α 2 , ∂z ∂z ∂z ∂z
(2.95)
∂p2 ∂p ∂α ∂Δp21 = + Δp21 1 + α1 . ∂z ∂z ∂z ∂z
(2.96)
Now we can write the specific form of the following general terms of the momentum equation ... + α leγ z
∂ pl ∂γ ⎤ ⎡∂ + ( pl − pm − δ lσ lmκ l − Δpmlσ ) ⎢ (α leγ z ) − δ l z ⎥ ∂z ∂z⎦ ⎣∂ z
− Δplwσ δ l
∂γ z + γ vα l ρl g cos ϕ ... ∂z
(2.97)
Note that both fields are continuous and therefore δ l = 1 and the effect of the surface tension is neglected ... + α1 γ z
∂ p1 ∂α ∂γ − γ z ( Δp21 + Δp12σ ) 1 − Δp1wσ z ... ∂z ∂z ∂z
(2.98)
... + α 2 γ z
∂ p2 ∂α ∂γ + γ z ( Δp21 − Δp12σ ) 2 − Δp2wσ z ... ∂z ∂z ∂z
(2.99)
Substituting the field pressures we finally obtain ⎡ ∂p ∂Δp21 ∂α ⎤ ∂γ ... + γ z ⎢α1 − α1 α 2 − (α 2 Δp21 + Δp12σ ) 1 ⎥ − Δp1wσ z ... ∂z ∂z ⎦ ∂z ⎣ ∂z
⎡ ∂p ∂Δp21 ∂α ... + γ z ⎢α 2 + α1α 2 − (α1 Δp21 − Δp12σ ) 1 ∂ z ∂ z ∂z ⎣
(2.100)
⎤ wσ ∂γ z ⎥ − Δp2 ∂ z ... (2.101) ⎦
Assuming that the change of the densities contributes much less to the change of
∂Δp21 ∂Δp21 ∂α1 H ∂α ≈ = g ( ρ1 − ρ 2 ) 1 ∂z ∂α1 ∂ z 2 ∂z
(2.102)
then the change of the local volume fraction becomes ⎡ ∂p ∂α ⎤ ∂γ ... + γ z ⎢α1 − gH α1α 2 ( ρ1 − ρ 2 ) 1 ⎥ − Δp1wσ z ... ∂z ⎦ ∂z ⎣ ∂z
(2.103)
⎡ ∂p ∂α ... + γ z ⎢α 2 + gHα1α 2 ( ρ1 − ρ 2 ) 1 ∂ z ∂z ⎣
(2.104)
⎤ wσ ∂γ z ⎥ − Δp2 ∂ z ... ⎦
2.7 Non-equal bulk and boundary layer pressures
69
As a plausibility check note that for α1 → 0 or α 2 → 0 the term containing the derivative of the volume fraction of velocity field 1 converges to zero and the momentum equations take the expected form. The sum of the two momentum equations gives ... + γ z
∂p ∂γ − ( Δp1wσ + Δp2wσ ) z ... ∂z ∂z
(2.105)
Note that the gravitational force is already taken into account in Eqs. (2.103) and (2.104) and there is no need for an additional term γ vα l ρl g cos ϕ . Stability criteria for the stratified flow can be obtained from the eigen values analysis. For simplicity assuming incompressible flow the mass and momentum equations of stratified flow in a straight pipe with constant cross section are
α1
∂w1 ∂w ∂α + α 2 2 + ( w1 − w2 ) 1 = 0 ∂z ∂z ∂z
(2.106)
∂α1 ∂w ∂α1 + α1 1 + w1 =0 ∂τ ∂z ∂z
(2.107)
α ( ρ − ρ1 ) ∂α1 ∂ w1 ∂w 1 ∂p + w1 1 + + gH 2 2 =0 ∂τ ∂z ρ1 ∂z ρ1 ∂z
(2.108)
α ( ρ − ρ1 ) ∂α1 ∂ w2 ∂w 1 ∂p + w2 2 + − gH 1 2 =0 ∂τ ∂z ρ 2 ∂z ρ2 ∂z
(2.109)
or in matrix notation
⎛0 ⎜ ⎜0 ⎜0 ⎜⎜ ⎝0
0 1 0 0
⎛ 0 ⎜ 0 0⎞ ⎛ p ⎞ ⎜ 0 ⎟ ⎜ ⎟ 0 0 ⎟ ∂ ⎜ α1 ⎟ ⎜ 1 +⎜ 1 0 ⎟ ∂τ ⎜ w1 ⎟ ⎜ ρ1 ⎟ ⎜ ⎟ ⎜ 0 1 ⎟⎠ ⎜⎝ w2 ⎟⎠ ⎜ 1 ⎜ρ ⎝ 2
(w
1
− w2 )
α1 α 2 ⎞
w1
α1
α 2 ( ρ 2 − ρ1 ) w1 ρ1 α ( ρ − ρ1 ) − gH 1 2 0 ρ2 gH
⎟ 0⎟ ⎛ p⎞ ⎟ ⎜α ⎟ ∂ ⎜ 1⎟ 0 ⎟ ⎜ ⎟ = 0. ⎟ ∂z w1 ⎟ ⎜⎜ ⎟⎟ ⎟ ⎝ w2 ⎠ w2 ⎟ ⎠ (2.110)
If you are not familiar with the analysis of the type of a system of partial differential equations by first computing the eigenvalues, eigenvectors and canonical forms it is recommended first to read Chapter 11 before continuing here. The eigenvalues are defined by the characteristics equations
70
2 Momentums conservation
⎛ 0 ⎜ ⎜ 0 ⎜ ⎜ 1 ⎜ ρ1 ⎜ ⎜ 1 ⎜ρ ⎝ 2
(w
1
− w2 )
α1
w1 − λ
α1
α 2 ( ρ 2 − ρ1 ) w1 − λ ρ1 α ( ρ − ρ1 ) − gH 1 2 0 ρ2 gH
α2 ⎞
⎟ ⎟ ⎟ 0 ⎟=0 ⎟ ⎟ ⎟ w2 − λ ⎟ ⎠ 0
(2.111)
or 2 α1 α α w1 − w2 ) ( w2 − λ ) − 1 ( w1 − λ ) ( w2 − λ ) − 2 ( w1 − λ ) ( ρ1 ρ1 ρ2
+ gH
α1α 2 ( ρ 2 − ρ1 ) =0 ρ1 ρ 2
(2.112)
or
α α ( ρ − ρ1 ) ⎛ α1 α 2 ⎞ 2 ⎛ α1 α2 ⎞ α α w1 ⎟ λ + 1 w22 + 2 w12 − gH 1 2 2 = 0. ⎜ + ⎟ λ − 2 ⎜ w2 + ρ ρ ρ ρ ρ ρ ρ1 ρ 2 ⎝ 1 2 ⎠ ⎝ 1 2 ⎠ 1 2 (2.113) This equation is in fact consistent with the gravity long wave theory of MilneThomson (1968). There are two eigen values
α1 α w2 + 2 w1 ± ρ1 ρ2 λ1,2 =
⎛ α1 α2 ⎞ w1 ⎟ ⎜ w2 + ρ2 ⎠ ⎝ ρ1
2
α α ( ρ − ρ1 ) ⎞ ⎛α α ⎞⎛α α − ⎜ 1 + 2 ⎟ ⎜ 1 w22 + 2 w12 − gH 1 2 2 ⎟ ρ ρ ρ ρ ρ1 ρ 2 ⎝ 1 2 ⎠⎝ 1 2 ⎠ ⎛ α1 α 2 ⎞ ⎜ + ⎟ ⎝ ρ1 ρ 2 ⎠ (2.114)
or after rearranging ⎡ ⎛α α ⎞ α1 α 2⎤αα w + 2 w ± ⎢ gH ( ρ 2 − ρ1 ) ⎜ 1 + 2 ⎟ − ( w1 − w2 ) ⎥ 1 2 ρ1 2 ρ 2 1 ρ ρ ⎝ 1 2 ⎠ ⎣ ⎦ ρ1 ρ 2 , λ1,2 = ⎛ α1 α 2 ⎞ ⎜ + ⎟ ⎝ ρ1 ρ 2 ⎠ (2.115)
which are real and different from each other if
2.7 Non-equal bulk and boundary layer pressures
( w1 − w2 )
2
⎛α α ⎞ < gH ( ρ 2 − ρ1 ) ⎜ 1 + 2 ⎟ . ⎝ ρ1 ρ 2 ⎠
71
(2.116)
In fact this is the Kelvin-Helmholtz stability criterion. If the above condition is satisfied the system describing the flow is hyperbolic. The violating the above condition results in the nature in a flow patterns that are different from the stratified one. Condition (2.116) is equivalent to Eq. (2.216) derived by Delhaye (1981), p. 90. Brauner and Maron included in 1992 the surface tension effect in their stability analysis and obtained
( w1 − w2 )
2
⎛α α ⎞ < H ⎜ 1 + 2 ⎟ ⎡⎣ g ( ρ 2 − ρ1 ) + σ 12 k 2 ⎤⎦ ⎝ ρ1 ρ 2 ⎠
- see Eq. (28a) in Brauner and Maron (1992) – which for neglected surface tension results in Eq. (2.116). Here k is the real wave number. 2.7.1.3 Stratified flow in horizontal or inclined pipes
Geometrical characteristics: The geometric flow characteristics for round pipes are non-linearly dependent on the liquid level, which makes the computation somewhat more complicated. Some important geometrical characteristics are specified here – compare with Fig. 2.9. The angle with the origin of the pipe axis defined between the upwards oriented vertical and the liquid-gas-wall triple point is defined as a function of the liquid volume fraction by the equation f (θ ) = − (1 − α 2 ) π + θ − sin θ cos θ = 0.
(2.117)
The derivative dθ π = dα1 2 sin 2 θ
(2.118)
will be used later. Having in mind that df = 2 sin 2 θ dθ
(2.119)
the solution with respect to the angle can be obtained by using the Newton iteration method as follows
θ = θ0 −
(1 − α 2 ) π − θ 0 + sin θ 0 cos θ 0 f0 = θ0 + df dθ 2 sin 2 θ 0
(2.120)
72
2 Momentums conservation
Per1
1
3
θ
Dh
1
δ 2F
2
( ρw)23
( ρw)32
ϕ
2
Per2
Fig. 2.9 Definition of the geometrical characteristics of the stratified flow
where index 0 stays for the previous guess. The iteration starts with an initial value of π / 2 Kolev (1977). Biberg proposed in 1999 accurate direct approximation 1/ 3
⎛ 3π ⎞ ⎟ ⎝ 2 ⎠
θ = πα 2 + ⎜
⎡1 − 2α 2 + α 21/ 3 − (1 − α 2 )1/ 3 ⎤ ⎣ ⎦
(2.120b)
with an error less then ±0.002rad or 1/ 3
⎛ 3π ⎞ ⎟ ⎝ 2 ⎠
θ = πα 2 + ⎜ −
⎡1 − 2α 2 + α 21/ 3 − (1 − α 2 )1/ 3 ⎤ ⎣ ⎦
{
}
1 2 α 2 (1 − α 2 )(1 − 2α 2 ) 1 + 4 ⎡α 22 + (1 − α 2 ) ⎤ , ⎣ ⎦ 200
(2.120c)
with an error less then ±0.00005rad . The perimeter of the pipe is then Per1w = π Dh ,
(2.121)
and the wetted perimeters for the gas and the liquid parts are Per1w = θ Dh ,
(2.122)
Per2w = (π − θ ) Dh .
(2.123)
The gas-liquid interface median is then b = Dh sin (π − θ ) = Dh sin θ ,
(2.124)
and the liquid level
δ 2F =
1 Dh (1 + cos θ ) . 2
(2.125)
2.7 Non-equal bulk and boundary layer pressures
73
The hydraulic diameters for the gas and the liquid for computation of the pressure drop due to friction with the wall are therefore Dh1 = 4α1 F / Per1w =
π α1 Dh θ
Dh 2 = 4α 2 F / Per2 w =
π
α 2 Dh
π −θ
(2.126) (2.127)
and the corresponding Reynolds numbers Re1 =
α1 ρ1w1 Dh π , η1 θ
(2.128)
Re2 =
α 2 ρ 2 w2 Dh π . η2 π −θ
(2.129)
Here F is the channel cross section and Per1 and Per2 are the wet perimeters of gas and film, respectively. If one considers the core of the flow the hydraulic diameter for computation of the friction pressure loss component at the gas-liquid interface is then Dh12 = 4α1 F / ( Per1w + b ) =
π α1 Dh θ + sin θ
(2.130)
and the corresponding Reynolds number Re1 =
α1 ρ1 w1 − w2 Dh π . η1 θ + sin θ
(2.131)
The gas-wall, liquid-wall and gas-liquid interfacial area densities are a1w =
Per1w 4α1 θ 4 = = , F Dh1 π Dh
(2.132)
a2 w =
Per2 w 4α 2 π − θ 4 , = = F Dh 2 π Dh
(2.133)
a12 =
b sin θ 4 = . F π Dh
(2.134)
Some authors approximated this relation for a smooth interface with a12 ≅
8
π Dh
α 2 (1 − α 2 ) ,
(2.135)
which in view of the accurate computation presented above is no more necessary. For the estimation of flow pattern transition criterion the following expression is some times required
74
2 Momentums conservation
dα 2 4 sin θ = . d δ 2 F Dh π
(2.136)
Gravitational (hydrostatic) pressure variation across the flow cross section of a horizontal pipe: In stratified flow the gravitation is a dominant force. The cross section averaged gas gravity pressure difference with respect to the interface is ⎛ sin 3 θ 1 ⎞ Δp12σ = ρ1 gDh ⎜ − cos θ ⎟ . ⎝ 3πα1 2 ⎠
(2.137)
The cross section averaged liquid gravity pressure difference with respect to the bottom of the pipe is ⎛ sin 3 θ 1 ⎞ Δp12σ = − ρ 2 gDh ⎜ + cos θ ⎟ . (2.138) 3 πα 2 ⎝ 2 ⎠ With respect to the interfacial pressure we have ⎛ sin 3 θ 1 ⎞ p1 = p12σ − Δp12σ = p12σ − ρ1 gDh ⎜ − cos θ ⎟ , ⎝ 3πα1 2 ⎠
(2.139)
⎛ sin 3 θ 1 ⎞ p2 = p12σ − Δp12σ = p12σ + ρ 2 gDh ⎜ + cos θ ⎟ , ⎝ 3πα 2 2 ⎠
(2.140)
which are in fact Eqs. (55) and (60) in Ransom et al. (1987) , p. 30. The averaged pressure in the cross section can be expressed as the cross section weighted averaged pressures inside the fields ⎡ sin 3 θ ⎤ 1 p = α1 p1 + α 2 p2 = α1 p12σ + α 2 p12σ + gDh ⎢ ( ρ2 − ρ1 ) + (α 2 ρ2 + α1 ρ1 ) cos θ ⎥ . 2 ⎣ 3π ⎦ (2.141)
Neglecting the surface tension we obtain ⎡ sin 3 θ ⎤ 1 p ≈ pσ + gDh ⎢ ( ρ2 − ρ1 ) + (α 2 ρ2 + α1 ρ1 ) cos θ ⎥ , 3 π 2 ⎣ ⎦
(2.142)
⎡ sin 3 θ ⎤ 1 pσ ≈ p − gDh ⎢ ( ρ2 − ρ1 ) + (α 2 ρ2 + α1 ρ1 ) cos θ ⎥ . 2 ⎣ 3π ⎦
(2.143)
or
The averaged pressure in the gas and in the liquid phase can then be expressed as a function of the system averaged pressure and geometrical characteristics: ⎡ sin 3 θ ⎤ 1 p1 = p − gDh ⎢ (α1 ρ2 + α 2 ρ1 ) + α 2 ( ρ2 − ρ1 ) cos θ ⎥ for α1 > 0 2 ⎣ 3πα1 ⎦ (2.144)
2.7 Non-equal bulk and boundary layer pressures
75
⎡ sin 3 θ ⎤ 1 p2 = p + gDh ⎢ (α1 ρ2 + α 2 ρ1 ) + α1 ( ρ2 − ρ1 ) cos θ ⎥ for α 2 > 0 2 ⎣ 3πα 2 ⎦ (2.145)
The check α1 p1 + α 2 p2 = p proves the correctness of the computation. The difference between both averaged pressures is then ⎡ sin 3 θ ⎛ ρ 2 ρ1 ⎞ ⎤ 1 Δp21 = p2 − p1 = gDh ⎢ + ⎟ + ( ρ 2 − ρ1 ) cos θ ⎥ ⎜ 2 ⎣ 3π ⎝ α 2 α1 ⎠ ⎦ for α1 > 0 and α 2 > 0
(2.146)
and therefore Eqs. (2.93-2.96) are valid also for a circular pipe. Assuming that the change of the densities contributes much less to the change of Δp21 then the change of the local volume fraction results in ∂Δp21 ∂Δp21 ∂α1 ≈ ∂α1 ∂α1 ∂ z ⎡ sin 3 θ = gDh ⎢ ⎣ 3π
⎛ ρ 2 ρ1 ⎞ ⎛ ρ 2 ρ1 ⎞ 1 π ⎤ ∂α1 + ⎟ cos θ − ( ρ 2 − ρ1 ) . ⎥ ⎜ 2 − 2 ⎟+⎜ α α α α 2 4 sin θ⎦ ∂z ⎝ 2 1 ⎠ ⎝ 2 1 ⎠ (2.147)
This equation was obtained by taking into account that θ is also an implicit function of α1 through Eq. (2.117) and using Eq. (2.118). Using Eqs. (2.156), (2.146), (2.137) and (2.138) and substituting into the momentum equations (2.100) and (2.101) we finally obtain ⎡ ∂p π gDh ∂α1 ⎤ ∂γ ... + γ z ⎢α1 + α1 α 2 ( ρ 2 − ρ1 ) − Δp1wσ z ... ⎥ 4 sin θ ∂ z ⎦ ∂z ⎣ ∂z
(2.148)
⎡ ∂p π gDh ∂α1 ⎤ wσ ∂γ z ... + γ z ⎢α 2 − α1α 2 ( ρ 2 − ρ1 ) ⎥ − Δp2 ∂ z ... ∂ z 4 sin θ ∂ z ⎣ ⎦
(2.149)
As plausibility check note that for α1 → 0 the term containing the derivative of the volume fraction of velocity field 1 converges to zero and the momentum equation for field 2 takes the expected form. The sum of the two equations is then ... + γ z
∂p ∂γ − ( Δp1wσ + Δp2wσ ) z ... ∂z ∂z
(2.150)
Comparing with the momentum equations for the rectangular channel we find inπ Dh π Dh2 stead of H the length = , which is the height of the rectangular channel 4 sin θ 4b
76
2 Momentums conservation
having the same cross section as the pipe and base equal to the gas-liquid median from Eq. (2.124). Therefore the stability condition for stratified flow in a pipe is
( w1 − w2 )
2
0 . The equations above contain derivatives (∂ pi /∂ ρi )T and (∂ pi /∂ T ) p . Only i
the derivatives (∂ρi / ∂ pi )T and ( ∂ρi /∂ T ) p of the simple components are known, i
so that the expressions for the mixture derivatives (3.53) through (3.56) in which the known component derivatives explicitly occur are easily obtained by replacing ⎛ ∂ pi ⎞ ⎛ ∂ρi ⎞ ⎜ ⎟ =1 ⎜ ⎟, ⎝ ∂ρi ⎠T ⎝ ∂ pi ⎠T ⎛ ∂ pi ⎞ ⎛ ∂ρi ⎞ ⎜ ⎟ = −⎜ ⎟ ⎝ ∂ T ⎠ ρi ⎝ ∂ T ⎠ pi
(3.57) ⎛ ∂ρi ⎞ ⎜ ⎟. ⎝ ∂ pi ⎠T
(3.58)
The final result is ⎛ ∂ρ ⎞ =1 ⎜ ⎟ ⎝ ∂ p ⎠T , all _ C ′s
⎡ mmax ⎛ ∂ρ m ⎞ ⎤ ⎢ ∑ Cm ⎜ ⎟ ⎥, ⎢⎣ m =1 ⎝ ∂ pm ⎠T ⎥⎦
mmax ⎛ ∂ρ ⎞ ⎛ ∂ρ ⎞ ⎛ ∂ρ m ⎞ = ∑ ⎜ ⎟ ⎜ ⎟ ⎜ ⎟ ⎝ ∂ T ⎠ p , all _ C ′s ⎝ ∂ p ⎠T , all _ C ′s m =1 ⎝ ∂ T ⎠ pm
(3.59)
⎛ ∂ρ m ⎞ ⎜ ⎟ , ⎝ ∂ pm ⎠T
(3.60)
⎡ ⎛ ∂ρi ⎞ ⎛ ∂ρ ⎞ ⎛ ∂ρ1 ⎞ ⎤ ⎛ ∂ρ ⎞ = −ρ ⎜ ⎢1 ⎜ ⎜ ⎟ ⎟ −1 ⎜ ⎟ ⎥ , (3.61) ⎟ ⎝ ∂ p ⎠T , all _ C ′s ⎢⎣ ⎝ ∂ pi ⎠T ⎝ ∂ p1 ⎠T ⎥⎦ ⎝ ∂ Ci ⎠ p ,T , all _ C ′s _ except _ Ci
for i = 2, mmax , and
⎛ ∂ρ ⎞ ⎛ ∂ρ ⎞ = ρ⎜ ⎜ ⎟ ⎟ ∂ C ⎝ ∂ p ⎠T , all _ C ′s ⎝ i ⎠ p ,T , all _ C ′s _ except _ Ci
⎛ ∂ρ1 ⎞ ⎜ ⎟ , ⎝ ∂ p1 ⎠T
(3.62)
for i = mmax + 1, imax . With these results Eq. (3.49) can be rewritten as a function of the pressure, temperature and concentrations
128
3 Derivatives for the equations of state
⎛∂ p ⎞ Cm ⎜ m ⎟ 1 ⎝ ∂ρ m ⎠T dpm = m dp + max ⎛∂ p ⎞ ⎛ ∂ρ m ⎞ Ci ⎜ i ⎟ ⎜ ⎟ ∑ i =1 ⎝ ∂ρi ⎠T ⎝ ∂ pm ⎠T ⎡ ⎢ ⎛ ∂ pm ⎞ ⎢ Cm +ρ ⎜ ⎟ ⎢ dCm − mmax ⎛ ∂ pi ⎞ ⎝ ∂ρ m ⎠T ⎢ ⎜ ⎟ Ci ∑ i =1 ⎝ ∂ρ i ⎠T ⎣⎢
⎡ ⎢ ⎢C ⎢ m ⎢ ⎢⎣
mmax
⎤ ⎛ ∂ρi ⎞ ⎥ ⎜ ⎟ pi ⎝ ∂ pi ⎠T ⎛ ∂ρ m ⎞ ⎥ − ⎜ ⎟ dT mmax ⎛ ∂ pi ⎞ ⎝ ∂ T ⎠ pm ⎥ ⎥ Ci ⎜ ⎟ ∑ ⎥⎦ i =1 ⎝ ∂ρ i ⎠T
⎛ ∂ρi ⎞
∑ ⎜⎝ ∂ T ⎟⎠ i =1
⎤ ⎥ ⎛ ∂ pi ⎞ ⎥ ⎜ ⎟ dCi ⎥ , ∑ i =1 ⎝ ∂ρ i ⎠T ⎥ ⎦⎥
mmax
(3.63)
or mmax ⎛∂ p ⎞ ⎛ ∂p ⎞ ⎛ ∂p ⎞ dpm = ⎜ m ⎟ dp + ⎜ m ⎟ dT + ∑ ⎜ m ⎟ dCi . ⎝ ∂T ⎠ p , all _ C ' s i = 2 ⎝ ∂ Ci ⎠ p ,T , all _ C ' s _ except _ C ⎝ ∂p ⎠T , all _ C ' s i (3.64)
Therefore
⎛ ∂pm ⎞ ⎜ ⎟ ⎝ ∂p ⎠T , all _ C ' s
⎛ ∂ρ ⎞ Cm ⎜ m ⎟ ⎝ ∂ pm ⎠T =m , max ⎛ ∂ρi ⎞ Ci ⎜ ∑ ⎟ i =1 ⎝ ∂ pi ⎠T
⎛ ∂pm ⎞ ⎜ ⎟ ⎝ ∂T ⎠ p , all _ C ' s
⎡ ⎢ 1 ⎢C = m ⎛ ∂ρ m ⎞ ⎢ ⎢ ⎜ ⎟ ⎝ ∂ pm ⎠T ⎣⎢
⎛ ∂ pm ⎞ ⎜ ⎟ ⎝ ∂ Ci ⎠ p ,T , all _ C ' s _ except _ Ci
(3.65)
⎤ ⎛ ∂ρi ⎞ ⎥ ⎜ ⎟ i =1 pi ⎝ ∂ pi ⎠T ⎛ ∂ρ m ⎞ ⎥ −⎜ , (3.66) ⎟ mmax ⎛ ∂ρi ⎞ ⎝ ∂ T ⎠ pm ⎥ ⎥ C ⎟ ∑ i ⎜ i =1 ⎝ ∂ pi ⎠T ⎦⎥
mmax
⎛ ∂ρi ⎞
∑ ⎜⎝ ∂ T ⎟⎠
⎡ ⎛ ∂ pi ⎞ ⎛ ∂ p1 ⎞ ⎢ ⎜ ⎟ −⎜ ⎟ ⎛ ∂ pm ⎞ ⎢ ⎝ ∂ρi ⎠T ⎝ ∂ρ1 ⎠T = ρ⎜ δ − C ⎟ ⎢ mi m mmax ⎛∂ p ⎞ ⎝ ∂ρ m ⎠T ⎢ Ci ⎜ i ⎟ ∑ i =1 ⎝ ∂ρi ⎠T ⎣⎢
⎤ ⎥ ⎥ . (3.67) ⎥ ⎥ ⎦⎥
3.2 Multi-component mixtures of miscible and non-miscible components
129
For the limiting case, where all of the components are taken to be perfect gases, we have ⎛ ∂ρi ⎞ 1 , ⎜ ⎟ = ∂ p R ⎝ i ⎠T iT
(3.68)
ρi ⎛ ∂ρi ⎞ ⎜ ⎟ =− , T ⎝ ∂ T ⎠ pi
(3.69)
and ⎛ ∂ρ1 ⎞ 1 , ⎜ ⎟ = ⎝ ∂ p1 ⎠T R1T
(3.70)
ρ1 ⎛ ∂ρ1 ⎞ ⎜ ⎟ =− . T ⎝ ∂ T ⎠ p1
(3.71)
Substituting in the equations defining the derivatives this yields ⎛ imax ⎞ ⎛ ∂ρ ⎞ 1 = 1 ⎜ T ∑ Ci Ri ⎟ = , ⎜ ⎟ ⎝ ∂ p ⎠T ,all _ C ′s ⎝ i =1 ⎠ RT
(3.72)
⎛ imax ⎞ ⎛ imax ⎞ ⎛ ∂ρ ⎞ = − ⎜ ∑ ρi Ri ⎟ ⎜ T ∑ Ci Ri ⎟ = − ρ / T , ⎜ ⎟ ⎝ ∂ T ⎠ p ,all _ C ′s ⎝ i =1 ⎠ ⎝ i =1 ⎠
(3.73)
⎛ ∂ρ ⎞ = − ρ ( Ri − R1 ) R , ⎜ ⎟ ⎝ ∂ Ci ⎠ p ,T ,all _ C ′s _ except _ Ci
(3.74)
imax
where R = ∑ Ci Ri . i =1
Now assume that the gas consists only of one i-th component, C1 = 0 , ρ1 = 0 and Ci = 1 . Using Eq. (3.50), this trivial case yields ⎛ ∂ρ ⎞ 1 = , ⎜ ⎟ ⎝ ∂ p ⎠T ,all _ C ′s RiT
(3.75)
⎛ ∂ρ ⎞ = − ρi / T , ⎜ ⎟ ⎝ ∂ T ⎠ p ,all _ C′s
(3.76)
and ⎛ ∂ρ ⎞ = 0. ⎜ ⎟ ⎝ ∂ Ci ⎠ p ,T ,all _ C′s _ except _ Ci
(3.77)
130
3 Derivatives for the equations of state
The derivatives are not defined in the case of missing miscible components because in this case Eq. (3.6) from which we started our derivation is no longer valid. In this case the Eq. (3.44) simplifies fortunately to 1
ρ
imax
=∑ i =1
Ci = 1 ρ C1,2,...,imax , p, T . ρ i ( p, T )
(
)
(3.78)
The derivatives are then obtained from the differential form imax
d ρ = ∑ Ci i =1
i ρ2 1 2 d ρ − ρ dCi ∑ i 2 ρi i =1 ρ i max
imax ⎡ imax ρ 2 ⎛ ∂ρ ⎞ ⎤ ⎡ imax ρ 2 ⎛ ∂ρ ⎞ ⎤ ⎛ 1 1 ⎞ = ⎢ ∑ Ci 2 ⎜ i ⎟ ⎥ dp + ⎢ ∑ Ci 2 ⎜ i ⎟ ⎥ dT − ρ 2 ∑ ⎜ − ⎟ dCi . (3.79) ρ1 ⎠ i = 2 ⎝ ρi ⎣⎢ i =1 ρi ⎝ ∂ p ⎠T ⎦⎥ ⎣⎢ i =1 ρi ⎝ ∂ T ⎠ p ⎥⎦
The final result is imax ⎛ ∂ρ ⎞ ρ 2 ⎛ ∂ρ ⎞ = ∑ Ci 2 ⎜ i ⎟ , ⎜ ⎟ ⎝ ∂ p ⎠T , all _ C ′s i =1 ρi ⎝ ∂ p ⎠T
(3.80)
imax ρ 2 ⎛ ∂ρ ⎞ ⎛ ∂ρ ⎞ = ∑ Ci 2 ⎜ i ⎟ , ⎜ ⎟ ⎝ ∂ T ⎠ p , all _ C ′s i =1 ρi ⎝ ∂ T ⎠ p
(3.81)
⎛ ∂ρ ⎞ ⎛ 1 1⎞ = −ρ 2 ⎜ − ⎟ , ⎜ ⎟ ⎝ ∂ Ci ⎠ p ,T , all _ C ′s _ except _ Ci ⎝ ρi ρ1 ⎠
(3.82)
If some of the non-miscible components are considered as non-compressible, their density derivatives are simply set to zero.
(
)
50405"Rctvkcn"fgtkxcvkxgu"kp"vjg"gswcvkqp"qh"uvcvg" T = T ϕ , p, C2,...,imax ." yjgtg" ϕ = s, h, e "
The next step is to define the partial derivatives in the linearized equation of state
(
T = T ϕ , p, C2,...,imax
namely
)
(3.83)
3.2 Multi-component mixtures of miscible and non-miscible components
131
imax ⎛ ∂T ⎞ ⎛ ∂T ⎞ ⎛ ∂T ⎞ dT = ⎜ dp + d ϕ + dCi , ⎜ ⎟ ∑ ⎟ ⎜ ⎟ i = 2 ⎝ ∂ Ci ⎠ p ,ϕ , all _ C ′s _ except _ C ⎝ ∂ p ⎠ϕ ,all _ C′s ⎝ ∂ϕ ⎠ p ,all _ C′s i (3.84)
where ϕ may be one of the following variables: specific entropy, specific enthalpy or specific internal energy,
ϕ = s, h, e. This is very important for the construction of a numerical algorithm within the framework of the entropy, or enthalpy or energy concepts. Remember that this kind of algorithm describes the behavior of the flow using the specific mixture property ϕ as elements of the dependent variable vector. As the increments δ p, δϕ and
δ Ci are known from the numerical integration of the system of PDEs governing the flow, it is then possible to compute the corresponding increment of the temperature T, and therefore, the particular thermo-physical and transport properties of the velocity field, which depend on T , p and Ci . Begin with the definition equation for the specific mixture property imax
ρϕ = ∑ ρiϕ i
(3.85)
i =1
or imax
ϕ = ∑ Ciϕ i .
(3.86)
i =1
After differentiating the definition equation for the specific mixture entropy (3.86), the following is obtained imax
imax
i =1
i =1
dϕ = ∑ Ci dϕi + ∑ ϕi dCi .
(3.87)
An arbitrary, but existing component is again denoted with 1, where the mass concentration of this component is uniquely defined by knowing the other concentrations imax
C1 = 1 − ∑ Ci
(3.88)
i =2
and therefore imax
dC1 = −∑ dCi . i =2
(3.89)
132
3 Derivatives for the equations of state
Note that this is the only criterion for selecting the above component denoted with 1. Its concentration is not calculated using the differential conservation equation, but merely using Eq. (3.88). Enthalpy concept ϕ = h . In this case Eq. (3.87) is imax
imax
i =1
i =1
dh = ∑ Ci dhi + ∑ hi dCi .
(3.90)
Replace in the above equation the differentials of the specific enthalpies using the caloric equations ⎛ ∂h ⎞ dhi = c pi dT + ⎜ i ⎟ dpi , ⎝ ∂ pi ⎠T
(3.91)
to obtain mmax imax ⎡ imax ⎛ ∂h ⎞ ⎛ imax ⎞ ⎛∂h ⎞ ⎤ dh = ⎜ ∑ Ci c pi ⎟ dT + ∑ Ci ⎜ i ⎟ dpi + ⎢ ∑ Ci ⎜ i ⎟ ⎥ dp + ∑ hi dCi . i =1 i =1 ⎝ i =1 ⎠ ⎝ ∂ pi ⎠T ⎣i =mmax +1 ⎝ ∂ p ⎠T ⎦ (3.92)
Replace dpi in the above equation by means of the differentiated equation of state pi = pi ( ρi , T ) = pi (Ci ρ , T ) , namely ⎛∂ p ⎞ ⎛∂ p ⎞ ⎛∂ p ⎞ dpi = ⎜ i ⎟ Ci d ρ + ρ ⎜ i ⎟ dCi + ⎜ i ⎟ dT , ∂ρ ∂ρ ⎝ ∂ T ⎠ ρi ⎝ i ⎠T ⎝ i ⎠T
(3.93)
to obtain ⎡⎛ imax ⎡ mmax ⎛ ∂ h ⎞ ⎛ ∂ p ⎞ ⎤ ⎞ mmax ⎛ ∂ h ⎞ ⎛ ∂ p ⎞ ⎤ dh = ⎢⎜ ∑ Ci c pi ⎟ + ∑ Ci ⎜ i ⎟ ⎜ i ⎟ ⎥ dT + ⎢ ∑ Ci2 ⎜ i ⎟ ⎜ i ⎟ ⎥ d ρ ⎢⎣ i =1 ⎢⎣⎝ i =1 ⎠ i =1 ⎝ ∂ pi ⎠T ⎝ ∂ T ⎠ ρi ⎥⎦ ⎝ ∂ pi ⎠T ⎝ ∂ρi ⎠T ⎥⎦ imax mmax ⎡ imax ⎛ ∂h ⎞ ⎛∂ p ⎞ ⎛∂h ⎞ ⎤ + ⎢ ∑ Ci ⎜ i ⎟ ⎥ dp + ∑ hi dCi + ∑ ρ Ci ⎜ i ⎟ ⎜ i ⎟ dCi . (3.94) ∂ p i =1 i =1 ⎠T ⎦ ⎝ ∂ pi ⎠T ⎝ ∂ρ i ⎠T ⎣i =mmax +1 ⎝
Substituting dC1 from Eq. (3.89) and d ρ from Eq. (3.47) into Eq. (3.94) yields imax ⎡ mmax ⎧⎪ ⎤ ⎤ ⎫⎪ ⎛ ∂ h ⎞ ⎡⎛ ∂ p ⎞ ⎛ ∂ p ⎞ ⎛ ∂ρ ⎞ dh = ⎢ ∑ Ci ⎨c pi + ⎜ i ⎟ ⎢⎜ i ⎟ + Ci ⎜ i ⎟ ⎜ + Ci c pi ⎥ dT ⎥ ⎬ ∑ ⎟ ⎢⎣ i =1 ⎩⎪ ⎥⎦ ⎝ ∂ pi ⎠T ⎣⎢⎝ ∂ T ⎠ ρi ⎝ ∂ρi ⎠T ⎝ ∂ T ⎠ p ,all _ C′s ⎦⎥ ⎭⎪ i=mmax +1
imax ⎧⎪ ⎡ mmax ⎛ ∂ h ⎞ ⎛ ∂ p ⎞ ⎤ ⎛ ∂ρ ⎞ ⎛ ∂ h ⎞ ⎫⎪ + ⎨ ⎢ ∑ Ci2 ⎜ i ⎟ ⎜ i ⎟ ⎥ ⎜ + ∑ Ci ⎜ i ⎟ ⎬ dp ⎟ ⎝ ∂ pi ⎠T ⎝ ∂ρi ⎠T ⎦⎥ ⎝ ∂ p ⎠T ,all _ C′s i= mmax +1 ⎝ ∂ p ⎠T ⎪⎭ ⎪⎩ ⎣⎢ i=1
3.2 Multi-component mixtures of miscible and non-miscible components imax ⎧ ⎫⎪ ⎡ mmax ⎛ ∂ h ⎞ ⎛ ∂ p ⎞ ⎤ ⎛ ∂ρ ⎞ ⎪ + ∑ ⎨ hi − h1 + ⎢ ∑ Ci2 ⎜ i ⎟ ⎜ i ⎟ ⎥ ⎜ ⎬ dCi ⎟ i=2 ⎪ ⎝ ∂ pi ⎠T ⎝ ∂ρi ⎠T ⎦⎥ ⎝ ∂ Ci ⎠ p ,T ,all _ C′s _ except _ Ci ⎪⎭ ⎣⎢ i=1 ⎩ mmax ⎡ ⎛ ∂h ⎞ ⎛∂ p ⎞ ⎛ ∂h ⎞ ⎛∂ p ⎞ ⎤ + ∑ ρ ⎢Ci ⎜ i ⎟ ⎜ i ⎟ − C1 ⎜ 1 ⎟ ⎜ 1 ⎟ ⎥ dCi , i =2 ⎝ ∂ p1 ⎠T ⎝ ∂ρ1 ⎠T ⎦⎥ ⎣⎢ ⎝ ∂ pi ⎠T ⎝ ∂ρi ⎠T
133
(3.95)
or imax ⎛ ∂h ⎞ ⎛∂h⎞ dh = c p dT + ⎜ dp + ∑ ⎜ dCi ⎟ ⎟ i = 2 ⎝ ∂ Ci ⎠ p ,T ,all _ C ′s _ except _ C ⎝ ∂ p ⎠T ,all _ C′s i
(3.96)
or
dT =
dh 1 ⎛ ∂ h ⎞ 1 − ⎜ dp − ⎟ c p c p ⎝ ∂ p ⎠T ,all _ C′s cp
imax
⎛ ∂h ⎞
i =2
⎝
∑⎜ ∂ C ⎟ i
⎠ p ,T ,all _ C ′s _ except _ Ci
dCi
(3.97)
where
cp =
=
mmax ⎧⎪ ⎛∂h ⎞ C c + Ci ⎨c pi + ⎜ i ⎟ ∑ ∑ i pi i = mmax +1 i =1 ⎝ ∂ pi ⎠T ⎩⎪ imax
⎡⎛ ∂ pi ⎞ ⎤ ⎫⎪ ⎛ ∂ pi ⎞ ⎛ ∂ρ ⎞ ⎢⎜ ⎥⎬ ⎟ ⎜ ⎟ + Ci ⎜ ⎟ ⎝ ∂ρi ⎠T ⎝ ∂ T ⎠ p ,all _ C′s ⎦⎥ ⎭⎪ ⎣⎢⎝ ∂ T ⎠ ρi
mmax ⎧⎪ (∂ hi ∂ pi )T Ci c pi + ∑ Ci ⎨c pi + (∂ρi ∂ pi )T i = mmax +1 i =1 ⎩⎪ imax
∑
⎡ ⎛ ∂ρ ⎞ ⎛ ∂ρ ⎞ ⎤ ⎫⎪ − ⎜ i ⎟ ⎥ ⎬ , (3.98) ⎢Ci ⎜ ⎟ ⎣⎢ ⎝ ∂ T ⎠ p ,all _ C′s ⎝ ∂ T ⎠ pi ⎦⎥ ⎭⎪
⎡ mmax 2 ⎛ ∂ hi ⎞ ⎛∂h⎞ ⎛ ∂ρ ⎞ =⎜ ⎢ ∑ Ci ⎜ ⎟ ⎜ ⎟ ⎟ ⎝ ∂ p ⎠T ,all _ C′s ⎝ ∂ p ⎠T ,all _ C′s ⎢⎣ i=1 ⎝ ∂ pi ⎠T
⎛ ∂h ⎞ = hi − h1 + Δhinp ⎜ ⎟ ⎝ ∂ Ci ⎠ p ,T ,all _ C ′s _ except _ Ci
imax ⎛ ∂ρi ⎞ ⎤ ⎛ ∂ hi ⎞ ⎜ ⎟ ⎥ + ∑ Ci ⎜ ⎟ , ⎝ ∂ pi ⎠T ⎥⎦ i= mmax +1 ⎝ ∂ p ⎠T (3.99)
(3.100)
134
3 Derivatives for the equations of state
and ⎡ ⎛ ∂h ⎞ Δhinp = ρ ⎢Ci ⎜ i ⎟ ⎢⎣ ⎝ ∂ pi ⎠T ⎡ mmax ⎛ ∂ h ⎞ + ⎢ ∑ Ci2 ⎜ i ⎟ ⎝ ∂ pi ⎠T ⎣⎢ i =1
⎛ ∂ρ i ⎞ ⎛ ∂ h1 ⎞ ⎜ ⎟ − C1 ⎜ ⎟ ⎝ ∂ pi ⎠T ⎝ ∂ p1 ⎠T
⎛ ∂ρ1 ⎞ ⎤ ⎜ ⎟ ⎥ ⎝ ∂ p1 ⎠T ⎥⎦
⎛ ∂ρi ⎞ ⎤ ⎛ ∂ρ ⎞ , ⎜ ⎟ ⎥⎜ ⎟ ⎝ ∂ pi ⎠T ⎦⎥ ⎝ ∂ Ci ⎠ p ,T ,all _ C′s _ except _ Ci
(3.101)
for i = 2, mmax and ⎡ mmax ⎛ ∂ h ⎞ Δhinp = ⎢ ∑ Ci2 ⎜ i ⎟ ⎢⎣ i=1 ⎝ ∂ pi ⎠T
⎛ ∂ρi ⎞ ⎤ ⎛ ∂ρ ⎞ ⎜ ⎟ ⎥⎜ ⎟ ⎝ ∂ pi ⎠T ⎥⎦ ⎝ ∂ Ci ⎠ p ,T ,all _ C′s _ except _ Ci
(3.102)
for i = mmax + 1, imax . Important: note that for non-existing miscible components the concentration denoted with “1” in Eq. (3.100) is the first non-miscible component. Equation (3.100) consists of two parts. In the case of a mixture consisting of perfect fluids, the second part is equal to zero, Δhilnp = 0 , because the enthalpies do not depend on the corresponding partial pressures. This also illustrates the meaning of the superscript np which stands for non-perfect fluid. From Eq. (3.97) we obtain the analytical expressions for the following derivatives ⎛ ∂T ⎞ 1 ⎛∂h⎞ =− ⎜ , ⎜ ⎟ ⎟ c p ⎝ ∂ p ⎠T ,all _ C′s ⎝ ∂ p ⎠h ,all _ C′s
(3.103)
1 ⎛∂T ⎞ = , ⎜ ⎟ ⎝ ∂ h ⎠ p ,all _ C′s c p
(3.104)
⎛ ∂T ⎞ 1 =− ⎜ ⎟ ∂ C c p ⎝ i ⎠ p ,h ,all _ C ′s _ except _ Ci
⎛ ∂h ⎞ . ⎜ ⎟ ⎝ ∂ Ci ⎠ p ,T ,all _ C ′s _ except _ Ci
(3.105)
Checking the above derivation: for a perfect gas mixture, and taking into account that ⎛ ∂ hi ⎞ ⎜ ⎟ = 0, ⎝ ∂ pi ⎠T
(3.106)
3.2 Multi-component mixtures of miscible and non-miscible components
⎛ ∂ h1 ⎞ ⎜ ⎟ = 0, ⎝ ∂ p1 ⎠T
135
(3.107)
we obtain imax
c p = ∑ Ci c pi ,
(3.108)
⎛∂h⎞ = 0, ⎜ ⎟ ⎝ ∂ p ⎠T ,all _ C ′s
(3.109)
i =1
and ⎛ ∂h ⎞ = hi − h1, ⎜ ⎟ ⎝ ∂ Ci ⎠ p ,T ,all _ C′s _ except _ Ci
(3.110)
which is the result expected. Energy concept ϕ = e . The derivation already presented for the specific enthalpy is formally identical with the derivation for the specific internal energy. One has to ⎛∂e ⎞ replace formally h with e and write instead c pi , ⎜ i ⎟ . ⎝ ∂ T ⎠ pi ,T Entropy concept ϕ = s. Then use the definition equation for the particular specific entropies T ρi dsi = ρi dhi − dpi
(3.111)
or divided by ρ TCi dsi = Ci dhi −
dpi
ρ
.
(3.112)
Substituting in this equation dhi using Eq. (3.91), summing the resulting imax equations and comparing the result with Eq. (3.87) we finally obtain imax ⎛ imax ⎞ dp Tds = ⎜ ∑ Ci dhi ⎟ − + T ∑ si dCi . i =1 ⎝ i=1 ⎠ ρ
(3.113)
Note that this is equivalent to the Gibbs equation for mixtures imax
Tds = ∑ d ( Ci hi ) − i =1
imax ⎛ imax ⎞ dp imax + T ∑ si dCi − ⎜ ∑ hi dCi ⎟ = dh − − ∑ ( h − Tsi ) dCi ρ ρ i =1 i i =1 ⎝ i =1 ⎠
dp
⎛ 1 ⎞ imax = de + pd ⎜ ⎟ − ∑ ( hi − Tsi ) dCi , ⎝ ρ ⎠ i =1
(3.114)
136
3 Derivatives for the equations of state
in which the hi − Tsi is the so called Gibbs potential for the single component. The entropy change due to the mixing process is dsmixing = −
1 imax ∑ ( hi − Tsi ) dCi . T i =1
(3.115)
Substituting dC1 from Eq. (3.89) into Eq. (3.115) results in dsmixing = −
1 imax ∑ ⎡ hi − h1 − T ( si − s1 )⎤⎦ dCi . T i =2 ⎣
(3.116a)
This result follows not from Eq. (3.114) directly but from the comparison of Eq. (3.114) with the energy conservation equation for the mixture. For a mixing process in which there is no total mass change in the system,
imax
∑Y M k =1
k
k
= const ,
Eq. (3.116a) can be rearranged using Eq. (3.17) dsmixing = −
1 imax ∑ Ci ⎡hi − h1 − T ( si − s1 )⎤⎦ d ln Yi . T i =2 ⎣
(3.116b)
Remember, that in case of mixing of perfect fluids the molar concentration is equivalent to the ratio of the partial pressure to the total pressure. Replace the differentials in Eq. (3.114) of the mixture specific enthalpy using the caloric equations (3.96) to obtain
ds =
cp T
dT +
imax ⎤ ⎛ Δhinp 1 ⎡ ⎛∂h⎞ − 1 dp + s − s + ⎢ρ ⎜ ⎥ ⎜ i 1 ∑ ⎟ ρT ⎣⎢ ⎝ ∂ p ⎠T , all _ C ′s ⎥⎦ T i=2 ⎝
⎞ ⎟dCi , ⎠
(3.117)
or
dT =
T ds − cp
⎛∂h⎞ −1 ⎟ ⎝ ∂ p ⎠T ,all _ C ′s
ρ⎜
ρcp
dp −
T cp
imax
⎛ ∂s ⎞
i =2
⎝
∑⎜ ∂ C ⎟ i
⎠ p ,T ,all _ C′s _ except _ Ci
dCi
(3.118)
where ⎛ ∂s ⎞ = si − s1 + Δsinp ⎜ ⎟ ∂ C ⎝ i ⎠ p ,T ,all _ C ′s _ except _ Ci
(3.119)
3.2 Multi-component mixtures of miscible and non-miscible components
137
where Δsinp =
Δhinp . T
(3.120)
Comparing Eq. (3.100) with Eq. (3.119) we realize that ⎛ ∂s ⎞ 1 ⎛ ∂h ⎞ = ⎜ . ⎜ ⎟ ⎟ ∂ C T ⎝ i ⎠ p ,T ,all _ C ′s _ except _ Ci ⎝ ∂ Ci ⎠ p ,T ,all _ C ′s _ except _ Ci
(3.121)
Like Eq. (3.100) Eq. (3.119) also consists of two parts. In the case of a mixture consisting of perfect fluids, the second part is equal to zero, Δsilnp = 0 , because the enthalpies do not depend on the corresponding partial pressures. This again illustrates the meaning of the superscript np which stands for non-perfect fluid. From Eq. (3.118) analytical expressions for the following derivatives are obtained
⎛ ∂T ⎞ 1 = ⎜ ⎟ ⎝ ∂ p ⎠ s ,all _ C′s ρ c p
⎡ ⎤ ⎛ ∂h⎞ ⎢1 − ρ ⎜ ⎥, ⎟ ⎝ ∂ p ⎠T ,all _ C′s ⎥⎦ ⎢⎣
T ⎛ ∂T ⎞ = , ⎜ ⎟ ⎝ ∂ s ⎠ p ,all _ C ′s c p
⎛ ∂T ⎞ T =− ⎜ ⎟ ∂ C c p ⎝ i ⎠ p ,s ,all _ C ′s _ except _ Ci
(3.122)
(3.123)
⎛ ∂s ⎞ . ⎜ ⎟ ⎝ ∂ Ci ⎠ p ,T ,all _ C ′s _ except _ Ci
(3.124)
Checking the above derivation: for a perfect gas mixture, and taking into account that ⎛ ∂ hi ⎞ ⎜ ⎟ = 0, ⎝ ∂ pi ⎠T
(3.125)
⎛ ∂ h1 ⎞ ⎜ ⎟ = 0, ⎝ ∂ p1 ⎠T
(3.126)
we obtain imax
c p = ∑ Ci c pi , i =1
(3.127)
138
3 Derivatives for the equations of state
⎛∂h⎞ = 0, ⎜ ⎟ ⎝ ∂ p ⎠T ,all _ C ′s
(3.128)
and ⎛ ∂s ⎞ = si − s1, ⎜ ⎟ ⎝ ∂ Ci ⎠ p ,T ,all _ C′s _ except _ Ci
(3.129)
which is the result expected. Thus, as already mentioned, after the required derivatives have been computed in Eq. (3.118) in the manner shown, the temperature difference can be easily computed: ⎛ ∂T ⎞ ∂T T − Ta = ⎜ ( p − pa ) + ⎛⎜ ⎞⎟ ( s − sa ) ⎟ ∂ p ∂ s ⎠ p ,all _ C′s ⎝ ⎝ ⎠ s ,all _ C ′s imax ⎛ ∂T ⎞ + ∑⎜ ( Ci − Cia ) , ⎟ i = 2 ⎝ ∂ Ci ⎠ p , s , all _ C ′s _ except _ C i
(3.130)
which corresponds to the increments for the pressure, entropy and concentrations. When the increments δ p , δ s and δ Ci are of considerable magnitude, it is better to take into account their non-linear dependencies in Eq. (3.118), imax ⎤ ⎛ ∂s ⎞ dT 1 ⎡ p = ⎢ ds − ∑ ⎜ dCi ⎥ + ⎟ T cp ⎢ i = 2 ⎝ ∂ Ci ⎠ p ,T , all _ C ′s _ except _ C ⎥⎦ ρTc p i ⎣
⎡ ⎤ dp ⎛ ∂h ⎞ ⎢1 − ρ ⎜ ⎥ ⎟ ⎝ ∂ p ⎠T ,all _ C′s ⎥⎦ p ⎢⎣ (3.131)
in the following way: substitute Δs* =
R=
imax ⎤ ⎛ ∂s ⎞ 1 ⎡ ⎢ s − sa − ∑ ⎜ ( C − Cia ) ⎥ ⎟ cp ⎢ i = 2 ⎝ ∂ Ci ⎠ p ,T , all _ C ′s _ except _ C ⎥⎦ i ⎣
p ρT
⎡ ⎤ ⎛∂h⎞ ⎢1 − ρ ⎜ ⎥ ⎟ ⎢⎣ ⎝ ∂ p ⎠T ,all _ C ′s ⎥⎦
(3.132)
(3.133)
and integrate to obtain ln (T Ta ) = Δs * + ( R c p ) ln ( p pa )
(3.134)
T = Ta e Δs* ( p pa )
(3.135)
or R cp
,
or for a quasi-linear relationship between temperature T and specific enthalpy h
3.2 Multi-component mixtures of miscible and non-miscible components
h = ha e Δs* ( p pa )
R cp
.
139
(3.136)
This is a canonical equation, as all thermodynamic properties of the gas mixture can be computed as a function of ( s, p, Ci ) by differentiating. For Δs* = 0 , T = Ta ( p pa )
R cp
.
(3.137)
Equation (3.130) is also useful for exact solution of the task T = ? if s, p and C2,3,...,imax are known. This can be performed by iterations (Newton method) as follows: the error for the previous step designated with superscript n is
(
)
Δs n = s − s n T n , p, C2,3,...,imax .
(3.138)
A new temperature T n+1 can be computed so as to obtain Δs n+1 = 0 , ⎛∂s ⎞ Δs n+1 − Δs n = − ⎜ (T n+1 − T n ) ⎟ ⎝ ∂ T ⎠ p ,all _ C ′s
(3.139)
or solving with respect to T n+1 and using Eq. (3.123)
{
(
)
}
)
}
T n+1 = T n 1 + ⎡⎣ s − s n T n , p, C2,3,...,imax ⎤⎦ c p
(3.140)
or, in more precise form
{
(
T n+1 = T n exp ⎡⎣ s − s n T n , p, C2,3,...,imax ⎤⎦ c p
(3.141)
50406"Ejgokecn"rqvgpvkcn" 3.2.4.1 Gibbs function
Neither the specific entropy, nor the specific enthalpy, nor the specific internal energy can be measured directly. The quantities that can be measured directly are pressure, temperature and concentrations. It is interesting to know whether Eq. (3.114) can be rewritten as a function of measurable variables f = f(p, T, C’s). The answer is yes and the result is d ( h − Ts ) =
dp
ρ
imax
− sdT + ∑ ( hi − Tsi ) dCi .
(3.142)
i =1
Introducing the so called free enthalpy or Gibbs function, g = h − Ts
Eq. (3.142) takes the remarkable form
(3.143)
140
3 Derivatives for the equations of state
dg =
dp
ρ
imax
− sdT + ∑ ⎡⎣ hi − h1 − T ( si − s1 ) ⎤⎦ dCi ,
(3.144)
i=2
with ⎛ ∂g ⎞ 1 = , ⎜ ⎟ ∂ p ρ ⎝ ⎠T , all _ C ' s
(3.145)
⎛ ∂g ⎞ = − s, ⎜ ∂T ⎟ ⎝ ⎠ p , all _ C ' s
(3.146)
⎛ ∂g ⎞ = hi − h1 − T ( si − s1 ) . ⎜ ⎟ ⎝ ∂ Ci ⎠ p , s , all _ C ′s _ except _ Ci
(3.147)
3.2.4.2 Definition of the chemical equilibrium
We identify a mixture that is at constant mixture pressure and temperature and that does not change the concentrations of its constituent as a mixture in chemical equilibrium. The remarkable property of Eq. (3.144) is that it provides a quantitative definition of the chemical equilibrium, namely: dg = 0 or g = const
(3.148)
Now consider the following chemical reaction imax
∑ n Symb i =1
i
i
= 0.
(3.149)
Here Symbi is the chemical identification symbol of the i-th species and ni is the stoichiometric coefficient equal to the number of kg-moles of each species that participates in the reaction. Usually ni < 0 for reactants that are reducing their mass and ni > 0 for products that are increasing their mass in the mixture. An
3.2 Multi-component mixtures of miscible and non-miscible components
141
example is −2 H 2 − O2 + 2 H 2 O = 0 . During the progress of a stoichiometric reaction all concentrations change not arbitrarily but so that the mass change of a single species is always proportional to ni M i . This is demonstrated as follows. The mass balance of the stoichiometric reaction (3.149) gives imax
∑n M i
i =1
i
= 0.
(3.150)
Applied to our last example: −2 × 2kg − 1× 32kg + 2 × 18kg = 0 . Note that ni M i is a natural constant belonging to the specific chemical reaction. On other hand, the sum of all mass sources per unit volume of the mixture is equal to zero, imax
∑μ i =1
i
= 0.
(3.151)
Selecting one arbitrary, but existing, mass source term designated with subscript 1 (usually having the minimal initial concentration if it is going to be consumed) and rearranging we have imax
∑μ
1
i =1
μi = 0. μ1
(3.152)
Equations (3.150) and (3.151) can then, and only then, be satisfied simultaneously if
μi =
ni M i μ1 . n1 M 1
(3.153)
Applying this to our example we obtain
μ H = μ H , μ O = ( 32 4 ) μ H = 8μ H , μ H O = − ( 36 4 ) μ H = −9μ H . 2
2
2
2
2
2
2
2
For the closed volume in which the chemical reaction happens the mass conservation for each species gives dCi / dτ = μi / ρ ,
(3.154)
and therefore dCi / dτ μi / ρ = , dC1 / dτ μ1 / ρ
(3.155)
or dCi = ni M i d
C1 = ni M i dξ , n1 M 1
(3.156)
142
3 Derivatives for the equations of state
where ξ is sometimes called in the literature the reaction progress variable. Substituting in Eq. (3.142) we obtain d ( h − Ts ) =
⎡ imax ⎤ − sdT + ⎢ ∑ ( hi − Tsi ) ni M i ⎥ dξ . ρ ⎣ i =1 ⎦
dp
(3.157)
For the case of chemical equilibrium, dg = 0, and at constant mixture pressure and temperature, dp = 0 and dT = 0, Eq. (3.157) reads imax
∑ ( h − Ts ) n M i
i =1
i
i
i
= 0.
(3.158)
This means that the function g = g (ξ ) possesses an extreme dg = 0 if Eq. (3.158) is fulfilled for all components. For a single component Eq. (3.142) results in d ( hi − Tsi ) =
dpi
ρi
− si dT .
(3.159)
For the case of a constant mixture temperature valid for each species, dT = 0, we have d ( hi − Tsi ) =
dpi
ρi
,
(3.160)
or after integration hi − Tsi = hi 0 − Tsi 0 +
pi
∫
p0
dpi′
ρi
.
(3.161)
For perfect gases Eq. (3.161) results in pi
hi − Tsi = hi 0 − Tsi 0 + TRi ∫
p0
⎛p ⎞ dpi′ = hi 0 − Tsi 0 + TRi ln ⎜ i ⎟ . pi ⎝ p0 ⎠
(3.162)
Substituting (3.162) in Eq. (3.158) we obtain imax
∑(h i =1
i0
imax ⎛p ⎞ − Tsi 0 ) M i ni + T ∑ ni M i Ri ln ⎜ i ⎟ = 0 , i =1 ⎝ p0 ⎠
(3.163)
or bearing in mind that R = M i Ri ,
(3.164)
3.2 Multi-component mixtures of miscible and non-miscible components
143
with R = 8314 J/(kg-mol K) being the universal gas constant, ni
⎛ pi ⎞ ⎟ = 0. i =1 ⎝ p0 ⎠
imax
imax
∑ ( hi 0 − Tsi 0 ) M i ni + TR ln ∏ ⎜ i =1
(3.165)
Introducing the so called pressure equilibrium factor imax
ni
⎛p ⎞ K p = ∏⎜ i ⎟ = i =1 ⎝ p0 ⎠ imax
∏p
i
i =1
p0n
ni
⎛ p⎞ =⎜ ⎟ ⎝ p0 ⎠
n i max
n
⎛ p⎞ n Yi i = ⎜ ⎟ K y , ∏ i =1 ⎝ p0 ⎠
(3.166)
where imax
n = ∑ ni ,
(3.167)
i =1
imax
K y = ∏ Yi i , n
(3.168)
i =1
and solving Eq. (3.165) with respect to Kp results in the expression well known in chemical thermodynamics defining mathematically the chemical equilibrium ⎡ 1 imax ⎤ K p = exp ⎢ − ( hi 0 − Tsi 0 ) ni M i ⎥ , ∑ ⎣ TR i =1 ⎦
(3.169a)
see for comparison in Warnatz et al. (2001) p. 42, or n
⎡ 1 imax ⎤ ⎛p ⎞ K y = ⎜ 0 ⎟ exp ⎢ − ( hi 0 − Tsi 0 ) ni M i ⎥ . ∑ ⎝ p⎠ ⎣ TR i =1 ⎦
(3.169b)
Therefore the pressure equilibrium factor for the particular reaction (3.149) can be analytically computed from the properties of the constituents. Another reaction generates another expression for the pressure equilibrium constant. The expression imax
imax
imax
i =1
i =1
i =1
ΔGch,0 = ∑ ( hi 0 − Tsi 0 ) ni M i = ∑ hi 0 ni M i − T ∑ si 0 ni M i ,
(3.170)
144
3 Derivatives for the equations of state
is called the molar Gibbs energy of the chemical reaction, Warnatz et al. (2001). The expressions imax
ΔH ch,0 = ∑ hi 0 ni M i ,
(3.171)
i =1
imax
ΔSch ,0 = ∑ si 0 ni M i ,
(3.172)
i =1
are called the enthalpy and entropy changes of the mixture due to the chemical reaction, respectively. With this Eq. (3.169b) can also be used in the form n
n
⎛ ΔH ch ,0 R imax si 0 ⎞ ⎛ ΔH ch ,0 ΔSch,0 ⎞ ⎛ p0 ⎞ ⎛p ⎞ K y = ⎜ 0 ⎟ exp ⎜ − + + ∑ ni ⎟, ⎟ = ⎜ ⎟ exp ⎜ − TR R ⎠ ⎝ p⎠ T Ri ⎠ i =1 ⎝ p⎠ ⎝ ⎝
(3.173) The choice of the reference pressure varies in the literature. Some authors set it to unity. In this case K p = p n K y . If the mixture pressure is selected as the reference pressure we have K p = K y . In this case ΔGch,0 = G ( p, T ) . In this case Eq. (3.162) in its integral form Δgi = TRi d ln ( pi p ) define a dimensionless property fi = pi p called the fugacity of component i in the solution of perfect fluids. The
general definition of the fugacity of component i is d ln f i = dg i (TRi ) . 3.2.4.3 Partial pressures of perfect fluid compounds in chemical equilibrium
Given the temperature T and the total pressure p of a mixture of jmax compounds that may react in a number imax of chemical reactions, then jmax
∑ n Symb j =1
ij
j
= 0 , for
i = 1, imax .
(3.174)
Here nij is the stoichiometric coefficient, < 0 for reactants and > 0 for products. We look for a solution of jmax molar concentrations
(
)
YT = Y1 , Y2 ,..., Y jmax ,
(3.175)
for which the system is in chemical equilibrium. We know from Dalton’s law that the system pressure is the sum of the partial pressures jmax
p = ∑ pj , j =1
Yi =
pj p
,
jmax
∑Y j =1
j
= 1.
(3.176)
3.2 Multi-component mixtures of miscible and non-miscible components
145
For each chemical reaction i we have the condition enforcing chemical equilibrium (3.174), jmax
K y ,i = ∏ Y j
n
i ⎛ ΔH ch ,0,i R jmax s j 0 ⎞ ⎛p ⎞ = ⎜ 0 ⎟ exp ⎜ − + ∑ nij ⎟ , for i = 1, imax. (3.177) ⎜ T R j ⎟⎠ j =1 ⎝ p⎠ ⎝
nj
j =1
where jmax
ni = ∑ nij .
(3.178)
j =1
Here R is the universal gas constant, Rj is the gas constant of the j-th component, jmax
ΔH ch ,0,i = ∑ h0 j nij Mj
(3.179)
j =1
and s0 j = s0 j R j
(3.180)
are the dimensionless reference entropies of the species j. Therefore we have jmax unknowns and 1+ imax equations (3.176) , (3.177). The missing jmax − imax − 1 are obtained from the proportions defined by the mass conservation equations jmax
∑n M j =1
ij
j
= 0 , for
i = 1, imax ,
(3.181)
resulting in dC j nij M ij
=
dCij* nij* M ij *
, for i = 1, imax , j = 1, jmax ,
(3.182)
where ij* refers for each i-reaction to the minimum but existing compound j, or dY j nij
=
dYij* nij *
, for i = 1, imax , j = 1, jmax .
(3.183)
Observe that in a chemical reaction for reactants that completely disappear we have dY j = −Y j , and for products that do not exist before the reaction but just originate we have dY j = Y j . Comparing this with the definition of the stoichiometric coefficient for which nij < 0 for reactants and > 0 for products, we have Yj nij
=
Yij * nij*
, for i = 1, imax , j = 1, jmax .
(3.184)
146
3 Derivatives for the equations of state
Significance. The theory of the chemical equilibrium has very wide-ranging applications that that are particularly important to physical chemists, metallurgists, solid-state physicists, and many engineers. It permits us to predict the equilibrium composition of a mixture of chemicals in an isolated vessel; it tells us how many phases of an alloy can exists together at any particular temperature and pressure etc. Examples of the application of these theoretical results are given in Chapter 6. 3.2.4.4 Phase equilibrium
The idea of chemical equilibrium is extendable to the so called phase equilibrium. If we apply the definitions of the Gibbs function, Eq. (3.144), to a mixture of two phases designated with ´ for the liquid and ´´ for its vapor we have dg =
dp
ρ
− sdT + ⎡⎣ h′′ − h′ − T ( s ′′ − s ′ ) ⎤⎦ dC ′ =
dp
ρ
− sdT − ⎡⎣ h′′ − h′ − T ( s ′′ − s ′ ) ⎤⎦ dC ′′.
(3.185) Because evaporation at a given constant pressure happens for pure liquids at a constant temperature the relation between the entropy change and the enthalpy change is s ′′ − s′ =
h′′ − h′ . T
(3.186)
This relation makes the Gibbs function of the mixture independent of the vapor mass concentration change at constant pressure and temperature. Therefore the equilibrium between the liquid and its vapor for any pressure and temperature is defined by dg = 0 .
(3.187)
Equation (3.186) can be rewritten in the form h′ − Ts ′ = h′′ − Ts ′′ . In other words, the equilibrium between the liquid and its vapor is defined by the equality of the specific Gibbs functions for each phase g ′ = g ′′.
(3.188)
If the liquid and the vapor initially in equilibrium are disturbed by dp resulting in dg ′ and dg ′′ , respectively, but in such a way that the resultant mixture is again in equilibrium we have g ′ + dg ′ = g ′′ + dg ′′ ,
or after using Eq. (3.188)
(3.189)
3.2 Multi-component mixtures of miscible and non-miscible components
dg ′ = dg ′′,
147
(3.190)
resulting in v′dp − s ′dT = v′′dp − s ′′dT
(3.191)
and finally dT v ′′ − v ′ = , dp s ′′ − s ′
(3.192)
derived for the first time by Benoit Pierre Emile Clapeyron in 1834. The significance of Clapeyron´s equation is in the description of the p-T line dividing the stable single phase state from the metastable state. This line is called the saturation line, Clausius, see in Elsner (1974) p. 327. Another derivation of Eq. (3.192) is obtained if one visualizes in the T-s and p-v diagrams the evaporation process as a Carnot cycle for infinitesimal values of dT and dp. Equalizing the work computed from both diagrams for the cycle results in Eq. (3.192). Using Eq. (3.186) the above relation reads dT 1 v ′′ − v ′ = , dp T h′′ − h′
(3.193)
Clapeyron’s equations also define the equilibrium state for the case of an equilibrium mixture of liquid and solid. In this case the corresponding properties along the solidification line have to be used. At low pressure for which the density difference between liquid and vapor is very large Eq. (3.193) simplifies to dT dp = v ′′ (T Δ h) . This equation has been known since 1828 in France as the August equation. Assuming the vapor behaves as a perfect gas results in dp p = ( Δh R ) dT T 2, where R is the vapor gas constant for the specific substance, (R. Clausius 1850). Assuming that the latent heat evaporation is a constant, integrating between an initial state 0 and a final state and rearranging results in a useful expression for extrapolation along the saturation line around the known state 0, ⎛ Δh T − T0 p − p0 = exp ⎜ p0 ⎝ RT T0
⎞ ⎟ −1. ⎠
(3.194)
A next step in the improvements of Eq. (3.194) is to be assumed that
(
)
Δh ≈ Δh0 + α T , resulting in dp p = Δh0 RT 2 dT + (α RT ) dT . After integrat-
ing we obtain
148
3 Derivatives for the equations of state
ln p = −
Δh0 α Δh α + ln T + 0 + ln p0 − ln T0 , RT R RT0 R
(3.195)
or ln p = A / T + B ln T + C ,
(3.196)
which is a form empirically proposed by G.R. Kirchhoff in 1858. Sometimes only the boiling properties at atmospheric pressure are known. Equation (3.194) allows us to compute the saturation temperature at a pressure different from the atmospheric pressure as long as the pressure remains far below the critical pressure. Equation (3.196) allows the three unknown constants to be fitted on three measured points. In fact the non-linear dependence of the group ( v′′ − v′ ) ( h′′ − h′ ) on the temperature requires a non-linear approximation of the saturation line of the type ln p = f (T ) e.g. Δh ≈ a1 + a2T + a3T 2 . In this case dp 1 ⎛ dT dT ⎞ = ⎜ a1 + a2 + a3 dT ⎟ . p R ⎝ T2 T ⎠
After integration between two pressure-temperature points we obtain R ln
⎛ 1 1⎞ p T = a1 ⎜ − ⎟ + a2 ln + a3 (T − T0 ) . p0 T T T 0 ⎝ 0 ⎠
The reference point can be also the critical point. In this case we have an accurate representation of the saturation line that satisfy the Clausius equation, is consistent with definition of the latent heat of vaporization and with the critical point of the fluid. This equation for is very useful in the practice. If the pressure is the known variable it is useful to approximate the above equation with T ′ ( p ) = 1 ⎡⎣ a + b ( log10 p ) ⎤⎦ as a first approximation and the solve it by iterations ⎡1 ⎤ p T T n +1 = a1 ⎢ − R ln − a2 ln − a3 (T − Tc ) ⎥ . pc Tc ⎣ Tc ⎦
3.2.4.5 Equilibrium of gas solution in water
Consider absorption of a gas i by water having temperature T2 . The pressure in the gas space is pi , gas . The partial pressure of the specie i in the water is pi , aq . The stoichiometric reaction is then −1× I gas + 1× I aq = 0 . The solution is reaching equilibrium. The equilibrium is described by Eq. (3.169a)
3.2 Multi-component mixtures of miscible and non-miscible components −1
149
1
⎛ pi , gas ⎞ ⎛ pi ,aq ⎞ ⎧ Mi ⎫ ⎡ hi 0, aq − hi 0, gas − T2 ( si 0, aq − si 0, gas ) ⎤ ⎬ . ⎜ ⎟ ⎜ ⎟ = exp ⎨− ⎣ ⎦ ⎝ p0 ⎠ ⎝ p0 ⎠ ⎩ RT2 ⎭
The equation can be rewritten as, pi , aq pi , gas
⎛ ⎜ p M = exp ⎜⎜− i p RT2 ⎜ ⎜ ⎝
⎞ ⎧ hi 0, aq (T2 , p0 ) − hi 0, gas (T2 , p0 ) ⎫⎟ ⎪⎪⎟ ⎪⎪ ⎨ ⎬⎟ . ⎪ ⎪⎟ ⎩⎪ −T2 ⎡⎣ si 0, aq (T2 , p0 ) − si 0, gas (T2 , p0 ) ⎤⎦ ⎭⎪ ⎟⎠
Considering the gas-water solution as a perfect fluid solution we obtain for the saturated molar concentration of specie i in the water the final expression
Yi , aq
⎧ ⎪⎪ M = Yi , gas exp ⎨− i ⎪ RT2 ⎩⎪
⎧ hi 0, aq (T2 , p0 ) − hi 0, gas (T2 , p0 ) ⎫⎫ ⎪⎪ ⎪⎪⎪⎪ ⎨ ⎬⎬ . ⎪ ⎪⎪ ⎩⎪ −T2 ⎡⎣ si 0, aq (T2 , p0 ) − si 0, gas (T2 , p0 ) ⎤⎦ ⎭⎪ ⎭⎪
Note that the specific enthalpies and entropies at the reference state for gas and gas dissolved in water have to differ in order to obtain a solution. Several values for aqueous gas solutions are tabulated in Oelkers et al. (1995). For known relations c pi (T ) and c pi , aq (T ) the above expression may be explicitly expressed as a function of temperature. If the exponential expression is only a temperature function the results reflects the empirically found Henry’s law. The Henry’s law says “The mass of non condensable gas dissolved in a liquid is proportional to the partial pressure of the gas above the liquid with which it is in equilibrium” Frequently in the chemical thermodynamic literature the solubility data are approximated by the Henry’s law in the following form p1i = k H ,2i ( p, T2 ) Y2i , ∞ .
Here the partial pressure of specie i in the gas phase is p1i and Y2i ,∞ is the saturation molar concentration of the same specie in the liquid.
150
3 Derivatives for the equations of state
⎛ ⎧hi 0, aq (T2 , p0 ) − hi 0, gas (T2 , p0 ) ⎫⎞⎟ ⎜ ⎪⎪⎟ M ⎪⎪ k H ,2i ( p, T2 ) = p exp ⎜− i ⎨ ⎬⎟ ⎜ RT2 ⎪ ⎪⎟ ⎜ ⎡ ⎤ − T s T , p − s T , p ⎜ ⎩⎪ 2 ⎣ i 0, aq ( 2 0 ) i 0, gas ( 2 0 ) ⎦ ⎭⎪⎟ ⎝ ⎠ is called the Henry’s coefficient. Note, that this idealization does not hold for many cases. There is a non linear pressure dependence of the Henry’s coefficient too especially for very high pressures. Knowing the Henry’s coefficient the molar concentration of the saturated solution is then Y2 i , ∞ = p1i k H ,2 i (T2 ) ,
and the corresponding mass concentration is C2 i , ∞ =
Y2i ,∞ M 2i ,∞
Y2i , ∞ M 2i , ∞ + (1 − Y2i , ∞ ) M H 2 O
.
If the mass concentration of the saturated solution is known the molar concentration is easily computed by Y2i , ∞ =
C2 i , ∞ M 2 i , ∞
C2i ,∞ M 2i ,∞ + (1 − C2i ,∞ ) M H 2O
.
(
)
50407"Rctvkcn"fgtkxcvkxgu"kp"vjg"gswcvkqp"qh"uvcvg" ρ = ρ p, ϕ , C2,...,imax ." yjgtg" ϕ = s, h, e "
Enthalphy concept ϕ = h . Finally, we replace dT in Eq. (3.96) using Eq. (3.47) and substitute ⎛ ∂ρ ⎞ ⎛ ∂ρ ⎞ 1 ⎛ ∂h⎞ ⎛ ∂ρ ⎞ =⎜ −⎜ , ⎜ ⎟ ⎟ ⎜ ⎟ ⎟ ∂ p ∂ p ∂ T c ⎠ p ,all _ C′s p ⎝ ∂ p ⎠T ,all _ C ′s ⎝ ⎠ h ,all _ C′s ⎝ ⎠T ,all _ C ′s ⎝
(3.197)
1 ⎛ ∂ρ ⎞ ⎛ ∂ρ ⎞ = ⎜ , ⎜ ⎟ ⎟ ⎝ ∂ h ⎠ p ,all _ C′s c p ⎝ ∂ T ⎠ p ,all _ C′s
(3.198)
⎛ ∂ρ ⎞ ⎜ ⎟ ⎝ ∂ Ci ⎠ p ,h ,all _ C ′s _ except _ Ci ⎛ ∂ρ ⎞ 1 ⎛ ∂h ⎞ ⎛ ∂ρ ⎞ , =⎜ −⎜ ⎟ ⎜ ⎟ ⎟ ∂ C ∂ T c ⎠ p ,all _ C ′s p ⎝ ∂ Ci ⎠ p ,T ,all _ C ′s _ except _ Ci ⎝ i ⎠ p ,T ,all _ C ′s _ except _ Ci ⎝
to obtain
(3.199)
3.2 Multi-component mixtures of miscible and non-miscible components
151
imax ⎛ ∂ρ ⎞ ⎛ ∂ρ ⎞ ⎛ ∂ρ ⎞ dρ = ⎜ dp + dh + dCi (3.200) ⎜ ⎟ ∑ ⎟ ⎜ ⎟ ⎝ ∂ h ⎠ p ,all _ C′s i = 2 ⎝ ∂ Ci ⎠ p ,h ,all _ C ′s _ except _ C ⎝ ∂ p ⎠h ,all _ C ′s i
which is the differential form for the equation of state
ρ = ρ ( p, h, C2,...,n
max
).
(3.201)
Energy concept ϕ = e . Again, the derivation already presented for the specific enthalpy is formally identical with the derivation for the specific internal energy. ⎛∂e ⎞ One has to replace formally h with e and write instead c pi , ⎜ i ⎟ . ⎝ ∂ T ⎠ pi ,T Entropy concept ϕ = s: We replace dT in Eq. (3.118) using Eq. (3.47) and substitute ⎛ ∂ρ ⎞ ⎛ ∂ρ ⎞ ⎛ ∂ρ ⎞ =⎜ −⎜ ⎜ ⎟ ⎟ ⎟ ∂ p ∂ p ⎝ ⎠ s ,all _ C′s ⎝ ⎠T ,all _ C ′s ⎝ ∂ T ⎠ p ,all _ C′s
⎛∂h⎞ −1 ⎟ ⎝ ∂ p ⎠T ,all _ C′s
ρ⎜
ρcp
=
1 ρ = , a2 κ p
(3.202) T ⎛ ∂ρ ⎞ ⎛ ∂ρ ⎞ =⎜ , ⎜ ⎟ ⎟ ⎝ ∂ s ⎠ p ,all _ C′s ⎝ ∂ T ⎠ p ,all _ C ′s c p
(3.203)
⎛ ∂ρ ⎞ ⎜ ⎟ ⎝ ∂ Ci ⎠ p ,s ,all _ C′s _ except _ Ci ⎛ ∂ρ ⎞ T ⎛ ∂s ⎞ ⎛ ∂ρ ⎞ , =⎜ −⎜ ⎟ ⎜ ⎟ ⎟ ⎝ ∂ Ci ⎠ p ,T ,all _ C′s _ except _ Ci ⎝ ∂ T ⎠ p ,all _ C′s c p ⎝ ∂ Ci ⎠ p ,T ,all _ C′s _ except _ Ci
(3.204)
to obtain
dρ =
imax ⎛ ∂ρ ⎞ dp ⎛ ∂ρ ⎞ +⎜ ⎟ ds + ∑ ⎜ dCi ⎟ 2 a ⎝ ∂ s ⎠ p ,all _ C ′s i = 2 ⎝ ∂ Ci ⎠ p , s ,all _ C ′s _ except _ C i
(3.205)
or dρ =
i ⎛ ∂ρ ⎞ ρ dp ⎛ ∂ρ ⎞ +⎜ ⎟ ds + ∑ ⎜ dCi , ⎟ κ p ⎝ ∂ s ⎠ p ,all _ C′s i = 2 ⎝ ∂ Ci ⎠ p ,s ,all _ C ′s _ except _ C max
i
(3.206)
152
3 Derivatives for the equations of state
which is the differential form for the equation of state
ρ = ρ ( p, s, C2,...,n
max
).
(3.207)
The value κ in Eq. (3.206) is defined as
ρ p
κ=
= cp
⎡ ⎤ ⎛∂h ⎞ ρ⎜ − 1⎥ ⎢ ⎟ ⎝ ∂ p ⎠T , all _ C ′s ⎥ ⎛ ∂ρ ⎞ ⎢⎛ ∂ρ ⎞ −⎜ ⎟ ⎢⎜⎝ ∂ p ⎟⎠ ⎥ ρcp ⎝ ∂ T ⎠ p , all _ C ′s T , all _ C ′s ⎢ ⎥ ⎢⎣ ⎥⎦
⎡ p ⎛ ∂ρ ⎞ ⎤ cp 1 T ⎛ ∂ρ ⎞ cp + ⎜ R⎥ = , ⎢ ⎜ ⎟ ⎟ ρ ⎝ ∂ T ⎠ p , all _ C ′s ⎥⎦ p ⎛ ∂ρ ⎞ cv ⎢⎣ ρ ⎝ ∂ p ⎠T , all _ C ′s ρ ⎜⎝ ∂ p ⎟⎠T , all _ C ′s
(3.208)
where R=
⎤ ⎛∂h⎞ p ⎡ ⎢1 − ρ ⎜ ⎥, ⎟ ρT ⎣⎢ ⎝ ∂ p ⎠T ,all _ C ′s ⎦⎥
(3.209)
T ⎛ ∂ρ ⎞ ⎛ ∂ρ ⎞ ⎜ ⎟ ⎜ ⎟ ⎤ ρ ⎝ ∂ T ⎠ p ,all _ C′s ⎛∂h⎞ ⎝ ∂ T ⎠ p ,all _ C′s 1 ⎡ cv = c p + R = cp + ⎢1 − ρ ⎜ ⎥ , (3.210) ⎟ ρ ⎢⎣ ⎛ ∂ρ ⎞ p ⎛ ∂ρ ⎞ ⎝ ∂ p ⎠T ,all _ C′s ⎥⎦ ⎜ ⎟ ρ ⎜⎝ ∂ p ⎟⎠T ,all _ C′s ⎝ ∂ p ⎠T ,all _ C′s
is generally valid for real gases as well for perfect gases. R is used to denote the pseudo gas constant, which is of course not in fact a constant for real gases. κ is the isentropic exponent for real gas mixtures. Note, that κ is not identical to the isentropic exponent for perfect gases in all cases. Only for mixtures of perfect gases does the pseudo gas constant R reduce to the gas constant for perfect mixtures because (∂ h ∂ p )T ,all _ C′s = 0 , and the expression above for the κ coefficient for real gases then reduces to the usual expression defining the isentropic exponent for perfect gas. For proof, we insert Eqs. (3.75), (3.76) into Eq. (3.208). The result is
κ=
cp cp − R
.
(3.211)
3.3 Mixture of liquid and microscopic solid particles of different chemical substances
153
An example for application of the general theory to an air-steam mixture is given in Appendix 3.1.
505"Okzvwtg"qh"nkswkf"cpf"oketqueqrke"uqnkf"rctvkengu" qh"fkhhgtgpv"ejgokecn"uwduvcpegu" This is a limiting case of the theory presented in the previous Section 3.2.
(
)
50503"Rctvkcn"fgtkxcvkxgu"kp"vjg"gswcvkqp"qh"uvcvg" ρ = ρ p, T , C2,...,imax "
For this particular case the derivatives in Eq. (3.47) given by Eqs. (3.80) through (3.82) reduce to ⎛ ∂ρ ⎞ ρ 2 ⎛ ∂ρ ⎞ = C1 2 ⎜ 1 ⎟ , ⎜ ⎟ ρ1 ⎝ ∂ p ⎠T ⎝ ∂ p ⎠T ,all _ C ′s
(3.212)
ρ 2 ⎛ ∂ρ ⎞ ⎛ ∂ρ ⎞ = C1 2 ⎜ 1 ⎟ , ⎜ ⎟ ρ1 ⎝ ∂ T ⎠ p ⎝ ∂ T ⎠ p ,all _ C ′s
(3.213)
⎛ ∂ρ ⎞ ⎛ 1 1⎞ = ρ2 ⎜ − ⎟. ⎜ ⎟ ⎝ ∂ Ci ⎠ p ,T ,all _ C ′s _ except _ Ci ⎝ ρ1 ρi ⎠
(3.214)
For missing inert components in the liquid,
imax
∑C i =2
i
= 0 , and ρ = ρ1 , Eqs. (3.213)
and (3.214) take values characteristic of the pure liquid. If the mixture consists solely of inert components,
imax
∑C i =2
i
= 1 , Eqs. (3.212) and (3.213) are not defined. In
this case δρ = 0 holds by definition. Another consequence of Eq. (3.214) is that the concentration change for a given component causes a density change if and only if the density differs from the density of the liquid, ρi ≠ ρ1 . In all other cases, ρi = ρ1 , the concentration change does not lead to a change in mixture density.
154
3 Derivatives for the equations of state
(
)
50504"Rctvkcn"fgtkxcvkxgu"kp"vjg"gswcvkqp"qh"uvcvg" T = T p, ϕ , C2,...,imax " yjgtg" ϕ = h, e, s "
For this simple case we obtain the following. Enthalpy concept ϕ = h : imax
c p = ∑ Ci c pi ,
(3.215)
⎛∂h⎞ ⎛∂h ⎞ = C1 ⎜ 1 ⎟ . ⎜ ⎟ ∂ p ⎝ ⎠T ,all _ C ′s ⎝ ∂ p ⎠T
(3.216)
i =1
Energy concept ϕ = e : imax ⎛ ∂e ⎞ ⎛ ∂e ⎞ = Ci ⎜ i ⎟ , ∑ ⎜ ⎟ ⎝ ∂T ⎠ p i =1 ⎝ ∂T ⎠ p
(3.217)
⎛ ∂e ⎞ ⎛∂e ⎞ = C1 ⎜ 1 ⎟ . ⎜ ⎟ ∂ p ⎝ ⎠T ,all _ C ′s ⎝ ∂ p ⎠T
(3.218)
Entropy concept ϕ = s : ⎛ ∂s ⎞ = si − s1. ⎜ ⎟ ⎝ ∂ Ci ⎠ p ,T ,all _ C′s _ except _ Ci
(3.219)
If the increments δ p , δ s and δ Ci are known, the temperature increments δ T can be computed from Eq. (3.118) in a manner analogous to that used in the preceding section. In the event that the increments δ p , δ s and δ Ci are large in magnitude, it is better to take into account the non-linear character of Eq. (3.118), as already demonstrated using Eqs. (3.131) through (3.141). If Δs* = 0 , Eq. (3.137) will once again result. For p = const Eq. (3.136) yields T = Ta eΔs*.
(3.220)
Usually R is significantly smaller for liquids or mixtures of liquids than for gases because of the marked differences in the densities. The temperature change caused by compression or decompression is thus observable after significant changes in pressure. Usually, the entropy change due to the heat and mass transfer causes a change in the liquid temperature T ≈ Ta e Δs*. The partial derivatives in the equation of state ρ = ρ ( p, s, C2,...,nmax ) are formally identical to those shown in Section 2.4, and are therefore not presented here.
3.4 Single-component equilibrium fluid
155
506"Ukping/eqorqpgpv"gswknkdtkwo"hnwkf" The restriction applied that one velocity field is in thermodynamic equilibrium is very strong and is, in fact, not necessary for an adequate description of multi-phase flow behavior. This assumption is, however, used in a number of applications to yield an approximate estimate of the order of magnitude of the processes. For this reason, this section briefly summarizes the derivatives calculated previously (see Kolev (1986b), for example) for equilibrium mixtures. Consider the properties of a fluid, assuming that it is in a thermodynamic equilibrium. Additional use will be made here of the following superscripts: ′ , saturated liquid, ′′ , saturated steam, ′′′ , saturated solid phase.
The properties on the saturation lines are functions of pressure only T ′ = T ′( p) , T ′′ = T ′′( p ) , ρ ′, ρ ′′, ρ ′′′, s′, s′′, s′′′ = f ( p ).
The same is valid for the properties d ρ ′ / dp, d ρ ′′ / dp, d ρ ′′′ / dp, ds′ / dp, ds′′ / dp, ds′′′ / dp = f ( p ) .
As a result, four regions are distinguished. 50603"Uwrgtjgcvgf"xcrqt"
For s > s′′ superheated steam exists with properties that are functions of the pressure and of the temperature
ρ = ρ ( p, T ) ,
(3.221)
⎛ ∂ρ ⎞ ⎛ ∂ρ ⎞ dρ = ⎜ ⎟ dp + ⎜ ⎟ dT , ⎝ ∂ T ⎠p ⎝ ∂ p ⎠T
(3.222)
⎛ ∂ρ ⎞ ⎜ ⎟ ⎝ ∂ p ⎠T
(3.223)
⎛ ∂ρ ⎞ ,⎜ ⎟ = f ( p, T ) , ⎝ ∂ T ⎠p
T = T ( p, s ) ,
(3.224)
dT ds R dp = + , T cp cp p
(3.225)
c p = c p ( p, T ) ,
(3.226)
156
3 Derivatives for the equations of state
R=
p ρT
⎡ ⎛∂h⎞ ⎤ ⎢1 − ρ ⎜ ⎟ ⎥ = R ( p, T ) , ⎝ ∂ p ⎠T ⎦ ⎣
ρ = ρ ( p, s ) , dρ =
(3.228)
dp ⎛ ∂ρ ⎞ ρ dp ⎛ ∂ρ ⎞ + ⎜ ⎟ ds = + ds , 2 a ⎝ ∂ s ⎠p κ p ⎜⎝ ∂ s ⎟⎠ p
⎛ ∂ρ ⎞ ⎛ ∂ρ ⎞ 1 ρ = =⎜ − 2 a κ p ⎝ ∂ p ⎟⎠T ⎜⎝ ∂ T ⎟⎠ p
(3.227)
⎛∂h⎞ ⎟ −1 ⎝ ∂ p ⎠T , ρcp
(3.229)
ρ⎜
⎛ ∂ρ ⎞ ⎛ ∂ρ ⎞ T , ⎜ ⎟ = ⎜ ⎟ ⎝ ∂ s ⎠ p ⎝ ∂ T ⎠ p cp
(3.230)
(3.231)
where c p = c p ( p, T ) ,
(3.232)
⎛∂h⎞ ⎜ ⎟ = f ( p, T ) . ⎝ ∂ p ⎠T
(3.233)
For a perfect gas, the state equations and the derivatives are easily computed. For real gases, it is assumed that analytical expressions for the state equation are known, and therefore analytical expressions for the derivatives required can easily be derived. 50604"Tgeqpuvtwevkqp"qh"gswcvkqp"qh"uvcvg"d{"wukpi"c"nkokvgf"" coqwpv"qh"fcvc"cxckncdng
For many new application fields the information regarding thermodynamic data is very limited but order of magnitude analyses are required for the practice. Therefore we present next a few brief examples of how to construct an approximate closed description of the thermodynamic state using only a few data points. 3.4.2.1 Constant thermal expansion coefficient and isothermal compressibility coefficient
The volumetric thermal expansion coefficient defined as 1 ⎛ dv ⎞ 1 ⎛ dρ ⎞ β = ⎜ ⎟ =− ⎜ , v ⎝ dT ⎠ p ρ ⎝ dT ⎟⎠ p
(3.234)
3.4 Single-component equilibrium fluid
157
and the isothermal coefficient of compressibility defined as 1 ⎛ dv ⎞ 1 ⎛ dρ ⎞ k =− ⎜ ⎟ = ⎜ ⎟ , v ⎝ dp ⎠T ρ ⎝ dp ⎠T
(3.235)
are measured experimentally. As an example solid Al2O3 has the properties
β = 5.0 × 10 −5 K −1 , k = 2.8 × 10−12 Pa −1 McCahan and Shepherd (1993). The reciprocal value of the isothermal compressibility is known as elasticity modulus, E = k −1 , having dimensions of pressure, Pa. Thus the equation of state ρ = ρ (T , p )
(3.236)
can be easily constructed starting with dρ
ρ
=
1 ⎛ ∂ρ ⎞ 1 ⎛ ∂ρ ⎞ dT + ⎜ ⎟ dp = − β dT + kdp , ρ ⎜⎝ ∂ T ⎟⎠ p ρ ⎝ ∂ p ⎠T
(3.237)
and integrating with respect to some reference temperature and pressure
ρ = ρ0 exp ⎡⎣ − β (T − T0 ) + k ( p − p0 ) ⎤⎦ .
(3.238)
For construction of the equation of state for specific enthalpy, entropy and internal energy the measured dependence on the specific heat as a function of temperature c p = c p (T )
(3.239)
is necessary. Using the differential relationships 1⎡ 1 ⎛ ∂ρ ⎞ ⎤ ⎢1 + T ⎜ ⎥ dp ρ ⎣⎢ ρ ⎝ ∂ T ⎟⎠ p ⎦⎥
dh = c p dT + = c p dT + ds =
cp T
(1 − β T ) exp ⎡ β
dT +
de = dh −
⎣
ρ0
1
ρ
(T − T0 ) − k ( p − p0 )⎤⎦ dp ,
(3.240)
cp 1 ⎛ ∂ρ ⎞ β exp ⎡⎣ β (T − T0 ) − k ( p − p0 ) ⎤⎦ dp , ⎟ dp = dT − 2 ⎜ ρ ⎝ ∂T ⎠p T ρ0 (3.241)
dp +
⎡ p ⎛ ∂ρ ⎞ ⎤ 1 ⎡ T ⎛ ∂ρ ⎞ p ⎛ ∂ρ ⎞ ⎤ d ρ = ⎢c p + 2 ⎜ ⎟ ⎥ dp ⎟ ⎥ dT + ⎢ ⎜ ⎟ + ⎜ ρ ⎢⎣ ρ ⎝ ∂ T ⎠ p ρ ⎝ ∂ p ⎠T ⎥⎦ ρ ρ ⎝ ∂ T ⎠ p ⎥⎦ ⎢⎣ p
2
⎛ p ⎞ 1 = ⎜ c p − β ⎟ dT + ( − β T + kp ) dp ρ ⎠ ρ ⎝
158
3 Derivatives for the equations of state
= c p dT − − +
p exp ⎡⎣ −k ( p − p0 ) ⎤⎦
ρ0
β exp ⎡⎣ β (T − T0 ) ⎤⎦ dT
βT exp ⎣⎡ β (T − T0 ) ⎦⎤ exp ⎣⎡ − k ( p − p0 ) ⎦⎤ dp ρ0 1
ρ0
exp ⎡⎣ β (T − T0 ) ⎤⎦ exp ⎡⎣ − k ( p − p0 ) ⎤⎦ kpdp ,
(3.242)
and integrating from some reference pressure and temperature we obtain T
h = h0 + ∫ c p (T ) dT −
(1 − β T ) exp ⎡ β k ρ0
T0
⎣
(T − T0 ) ⎤⎦ {exp ⎡⎣ − k ( p − p0 )⎤⎦ − 1} , (3.243)
T
s = s0 + ∫ T0
c p (T ) T
dT +
T
e = e0 + ∫ c p (T ) dT − T0
+
+
1 β exp ⎡⎣ β (T − T0 ) ⎤⎦ exp ⎡⎣ − k ( p − p0 ) ⎤⎦ − 1 , (3.244) k ρ0
{
p
ρ0
}
{
}
exp ⎡⎣ − k ( p − p0 ) ⎤⎦ exp ⎡⎣ β (T − T0 ) ⎤⎦ − 1
βT exp ⎡⎣ β (T − T0 ) ⎤⎦ exp ⎡⎣ − k ( p − p0 ) ⎤⎦ − 1 k ρ0
{
}
p
1
exp ⎡⎣ β (T − T0 ) ⎤⎦ ∫ exp ⎡⎣ − k ( p − p0 ) ⎤⎦ kpdp . ρ0 p0
(3.245)
Many authors enforce s = 0 for T = 0. This is called in some text books the third law of the thermodynamics Cordfunke and Konings (1990) as proposed by Max Plank in 1912. Note that the solution of the last integral is quite complicated
π ⎛1 ∫ exp ⎡⎣−k ( p − p )⎤⎦ kpdp = − 2k exp ⎜⎝ 4 k p
0
p0
2
⎞⎡ p02 ⎟ ⎢erf ⎠⎣
1 ⎛ ⎞ ⎜ −kp + kp0 ⎟ + erf 2 ⎝ ⎠
⎛1 ⎞⎤ ⎜ kp0 ⎟ ⎥ . ⎝2 ⎠⎦
(3.246) In addition from the equation of state in the form d ρ = − βρ
⎧⎪ ⎫⎪ ⎛ ∂ρ ⎞ T T ⎛ ∂ρ ⎞ ds + ⎨ ρ k − β 2 ⎬ dp = ⎜ ⎟ ds + ⎜ ⎟ dp , cp c ∂ s ⎝ ⎠p ⎝ ∂ p ⎠s ⎪⎭ p ⎩⎪
(3.247)
we obtain the derivatives
⎛ ∂ρ ⎞ ⎜ ⎟ = −T βρ / c p , ⎝ ∂ s ⎠p
(3.248)
3.4 Single-component equilibrium fluid
⎛ ∂ρ ⎞ 2 ⎜ ⎟ = ρk − T β / cp . ⎝ ∂ p ⎠s
159
(3.249)
The velocity of sound is therefore a=
1
ρ k − T β 2 / cp
.
(3.250)
The derivatives of the enthalpy, entropy and specific internal energy with respect to pressure and temperature follow immediately from Eqs. (3.228) through (3.242). ⎛∂h⎞ ⎜ ⎟ = cp, ⎝ ∂T ⎠p
(3.251)
⎛∂h⎞ (1 − β T ) exp ⎡ β T − T − k p − p ⎤ , ( ⎜ ⎟ = 0) 0 )⎦ ⎣ ( ρ0 ⎝ ∂ p ⎠T
(3.252)
cp ⎛∂s ⎞ ⎜ ⎟ = , ⎝ ∂T ⎠p T
(3.253)
⎛∂s ⎞ β ⎜ ⎟ = − exp ⎡⎣ β (T − T0 ) − k ( p − p0 ) ⎤⎦ , ρ0 ⎝ ∂ p ⎠T
(3.254)
βp ⎛ ∂e ⎞ exp ⎡⎣ β ( T − T0 ) − k ( p − p0 ) ⎤⎦ , ⎜ ⎟ = cp − ρ0 ⎝ ∂T ⎠p
(3.255)
⎛ ∂e ⎞ kp − β T exp ⎣⎡ β (T − T0 ) − k ( p − p0 ) ⎦⎤ . ⎜ ⎟ = ∂ p ρ0 ⎝ ⎠T
(3.256)
3.4.2.2 Properties known only at given a pressure as temperature functions Frequently in the literature there are measured properties at atmospheric pressure p0 given as temperature functions only
ρ p = ρ p (T ) ,
(3.257)
⎛ ∂ρ ⎞ ⎜ ⎟ = f (T ) , ⎝ ∂T ⎠p
(3.258)
0
0
or
160
3 Derivatives for the equations of state
c p = c p (T ) .
(3.259)
In addition the velocity of sound is measured a = a ( T ).
(3.260)
For practical application of this information the reconstruction of the properties in consistent data functions is required. Next we construct the analytical form of the density as a function of temperature and pressure. The density derivative with respect to pressure at constant temperature ⎛ dρ ⎞ 1 1 T 2 ⎜ ⎟ = 2 + T β / cp = 2 + 2 a ρ cp ⎝ dp ⎠T a
2
⎛ dρ ⎞ ⎜ ⎟ = f (T ) ⎝ dT ⎠ p
(3.261)
is obviously a function of the temperature only. Therefore the density can be computed by integrating analytically the equation of state ⎛ ∂ρ ⎞ ⎛ ∂ρ ⎞ dρ = ⎜ ⎟ dp . ⎟ dT + ⎜ ⎝ ∂T ⎠p ⎝ ∂ p ⎠T
(3.262)
The result is ⎛ ∂ρ ⎞ ⎟ ( p − p0 ) . ⎝ ∂ p ⎠T
ρ = ρp + ⎜ 0
(3.263)
Using the differential equations
dp 1 ⎛ ∂ρ ⎞ 1⎡ 1 ⎛ ∂ρ ⎞ ⎤ +T 2 ⎜ dp , (3.264) ⎢1 + T ⎜ ⎟ ⎥ dp = c p dT + ρ ⎢⎣ ρ ⎝ ∂ T ⎠ p ⎥⎦ ρ ρ ⎝ ∂ T ⎟⎠ p
dh = c p dT +
ds =
cp T
dT +
de = dh −
= c p dT +
1
ρ
1 ⎛ ∂ρ ⎞ dp , ρ 2 ⎜⎝ ∂ T ⎟⎠ p
dp +
(3.265)
⎡ p ⎛ ∂ρ ⎞ ⎤ 1 ⎡ T ⎛ ∂ρ ⎞ p ⎛ ∂ρ ⎞ ⎤ d ρ = ⎢c p + 2 ⎜ ⎥ dp ⎟ ⎥ dT + ⎢ ⎜ ⎟ + ⎜ ρ ρ ⎝ ∂ T ⎠ p ⎥⎦ ρ ⎢⎣ ρ ⎝ ∂ T ⎠ p ρ ⎝ ∂ p ⎟⎠T ⎥⎦ ⎢⎣ p
2
p ⎛ ∂ρ ⎞ ⎛ ∂ρ ⎞ dT + T ⎜ ⎟ ρ 2 ⎜⎝ ∂ T ⎟⎠ p ⎝ ∂T ⎠p ⎡
dp
⎤ ⎛ ∂ρ ⎞ ⎢ ρ p0 + ⎜ ⎟ ( p − p0 ) ⎥ ⎝ ∂ p ⎠T ⎣ ⎦
⎛ ∂ρ ⎞ pdp +⎜ , ⎟ 2 ∂ p ⎝ ⎠T ⎡ ⎤ ⎛ ∂ρ ⎞ ⎢ ρ p0 + ⎜ ⎟ ( p − p0 ) ⎥ ⎝ ∂ p ⎠T ⎣ ⎦
2
(3.266)
3.4 Single-component equilibrium fluid
keeping in mind that p
∫
p0
dp 1 ρ , = ln ρ p0 ⎛ ∂ρ ⎞ ⎛ ∂ρ ⎞ ρ p0 + ⎜ ⎟ ( p − p0 ) ⎜ ⎟ ⎝ ∂ p ⎠T ⎝ ∂ p ⎠T
p
dp
∫⎡
p0
⎤ ⎛ ∂ρ ⎞ ⎢ ρ p0 + ⎜ ⎟ ( p − p0 ) ⎥ ⎝ ∂ p ⎠T ⎣ ⎦
⎛ ∂ρ ⎞ ⎜ ⎟ ⎝ ∂ p ⎠T
p
=
p − p0
ρp ρ
,
(3.268)
0
pdp
∫⎡
p0
2
(3.267)
⎤ ⎛ ∂ρ ⎞ ⎢ρp + ⎜ ⎟ ( p − p0 ) ⎥ ⎝ ∂ p ⎠T ⎣ ⎦
2
0
=
⎧⎪⎛ 1 ⎤ ⎛ ∂ρ ⎞ 1 ⎞⎡ ρ ⎫⎪ ⎟⎟ ⎢ ρ p0 − ⎜ ⎨⎜⎜ − ⎬, ⎟ p0 ⎥ + ln ρ p0 ⎪⎭ ⎛ ∂ρ ⎞ ⎩⎪⎝ ρ ρ p0 ⎠ ⎣ ⎝ ∂ p ⎠T ⎦ ⎜ ⎟ ⎝ ∂ p ⎠T 1
(3.269)
and integrating from some reference pressure and temperature we obtain T
h = h0 + ∫ c p (T ) dT + T0
1 ⎛ ∂ρ ⎞ ⎜ ⎟ ⎝ ∂ p ⎠T
ln
ρ ⎛ ∂ρ ⎞ p − p0 +T ⎜ , ⎟ ρp ⎝ ∂ T ⎠p ρp ρ 0
dp ⎛ ∂ρ ⎞ dT + ⎜ ⎟ 2 ∂ T ⎝ ⎠ ⎤ p ⎡ T0 ⎛ ∂ρ ⎞ ρ + p − p ⎢ p0 ⎜ ⎟ ( 0 )⎥ ⎝ ∂ p ⎠T ⎣ ⎦ T c ⎛ ∂ρ ⎞ p − p0 = s0 + ∫ p dT + ⎜ , ⎟ T ⎝ ∂ T ⎠ p ρ p0 ρ T0 T
s = s0 + ∫
(3.270)
0
cp T
T
T
T0
T0
e = e0 + ∫ c p dT + p ∫
(3.271)
1 ⎛ ∂ρ ⎞ dT ρ 2 ⎜⎝ ∂ T ⎟⎠ p
1 ⎛ ∂ρ ⎞ p − p0 +T ⎜ + ⎟ ∂ T ρ ρ ⎝ ⎠ p p0 ⎛ ∂ρ ⎞ ⎜ ⎟ ⎝ ∂ p ⎠T
⎧⎪⎛ 1 ⎤ ⎛ ∂ρ ⎞ 1 ⎞⎡ ρ ⎫⎪ ⎟⎟ ⎢ ρ p0 − ⎜ ⎨⎜⎜ − ⎬. ⎟ p0 ⎥ + ln ρ p0 ⎭⎪ ⎝ ∂ p ⎠T ⎦ ⎩⎪⎝ ρ ρ p0 ⎠ ⎣ (3.272)
161
162
3 Derivatives for the equations of state
Again the derivatives are
⎛∂h⎞ ⎜ ⎟ = cp , ⎝ ∂T ⎠p
(3.273)
⎛∂h⎞ 1⎡ 1 ⎛ ∂ρ ⎞ ⎤ ⎥, ⎜ ⎟ = ⎢1 + T ⎜ ρ ⎝ ∂ T ⎟⎠ p ⎥⎦ ⎝ ∂ p ⎠T ρ ⎢⎣
(3.274)
cp ⎛∂s ⎞ , ⎜ ⎟ = ⎝ ∂T ⎠p T
(3.275)
⎛ ∂s ⎞ 1 ⎛ ∂ρ ⎞ ⎜ ⎟ = 2⎜ ⎟ , ∂ p ρ ⎝ ∂ T ⎠p ⎝ ⎠T
(3.276)
⎡ p ⎛ ∂ρ ⎞ ⎤ ⎛ ∂e ⎞ ⎥, ⎜ ⎟ = ⎢c p + 2 ⎜ ρ ⎝ ∂ T ⎟⎠ p ⎦⎥ ⎝ ∂ T ⎠ p ⎣⎢
(3.277)
⎛ ∂e ⎞ ⎛ ∂ρ ⎞ ⎤ 1 ⎡ ⎛ ∂ρ ⎞ ⎜ ⎟ = 2 ⎢T ⎜ ⎟ ⎥. ⎟ + p⎜ ⎝ ∂ p ⎠T ρ ⎢⎣ ⎝ ∂ T ⎠ p ⎝ ∂ p ⎠T ⎦⎥
(3.278)
3.4.2.3 Constant density approximations For an idealized case of constant density we obtain
ρ = ρ0 ,
(3.279) T
h = h0 + ∫ c p (T ) dT ,
(3.280)
T0
T
s = s0 + ∫ T0
cp T
dT ,
(3.281)
T
e = e0 + ∫ c p dT .
(3.282)
T0
For zero initial values at the reference point we have
e ≡ h.
(3.283)
Within this concept internal pressure for the incompressible field is not defined. Submerged in fluid the interface averaged stress is the pressure of the system.
3.4 Single-component equilibrium fluid
163
3.4.2.4 Perfect gas approximations For a perfect gas defined by p = RT ρ
(3.284)
with constant specific heat at constant pressure the differential Eqs. (3.241, 3.242, 3.243) result in 1⎡ 1 ⎛ ∂ρ ⎞ ⎤ ⎢1 + T ⎜ ⎥ dp = c p dT , ρ ⎢⎣ ρ ⎝ ∂ T ⎟⎠ p ⎥⎦
(3.285)
dT 1 ⎛ ∂ρ ⎞ dT dp + 2⎜ dp = c p −R , T ρ ⎝ ∂ T ⎟⎠ p T p
(3.286)
dh = c p dT + ds = c p
de = dh −
1
ρ
dp +
⎡ p ⎛ ∂ρ ⎞ ⎤ 1 ⎡ T ⎛ ∂ρ ⎞ p ⎛ ∂ρ ⎞ ⎤ d ρ = ⎢c p + 2 ⎜ ⎥ dp ⎟ ⎥ dT + ⎢ ⎜ ⎟ + ⎜ ρ ρ ⎝ ∂ T ⎠ p ⎦⎥ ρ ⎢⎣ ρ ⎝ ∂ T ⎠ p ρ ⎝ ∂ p ⎟⎠T ⎦⎥ ⎣⎢ p
2
= ( c p − R ) dT = cv dT ,
(3.287)
h = c p (T − T0 ) ,
(3.288)
e = cv (T − T0 ) ,
(3.289)
or
s = c p ln
T p − R ln . T0 p0
(3.290)
Obviously for p ≤ 0 ln p p0 and consequently s is not defined. With other words: Gases does not exist at zero pressure and gases do not possess tension state. Modeling multi-phase flows using multi-velocity field models offers the possibility of postulating one of the mixtures described above for each of the fields. The combination of velocity fields and an appropriate number of components in each field make it possible to describe a large variety of flows observed in nature and in the field of engineering. 50605"Xcrqt/nkswkf"okzvwtg"kp"vjgtoqf{pcoke"gswknkdtkwo" For s′ < s < s ′′ a saturated liquid coexists with its saturated vapor in thermodynamic equilibrium after having enough time for relaxation. The mass concentration of the vapor in the equilibrium mixture is denoted with C”. The mass concentration is used to generate the following definitions for the specific entropy and the density of the mixtures
164
3 Derivatives for the equations of state
s = C ′′s′′ + (1 − C ′′ ) s′,
(3.291)
C ′′ 1 − C ′′ + . ρ ′′ ρ′
(3.292)
1
ρ
=
As p (and therefore s′′ , s′ ) and s are known, this permits easy computation of C ′′ from Eq. (3.291), i.e., C ′′ =
s − s′ . s′′ − s′
(3.293)
As C ′′ and the pressure are known, the density is uniquely defined by Eq. (3.292), i.e.,
ρ = ρ ( p, s ) ,
(3.294)
or dρ =
dp ⎛ ∂ρ ⎞ + ⎜ ⎟ ds. a2 ⎝ ∂ s ⎠ p
(3.295)
The derivatives can be easily obtained after differentiating Eqs. (3.291) and (3.292) and after eliminating dC”. The result is
ρ 2 ρ ′ − ρ ′′ ⎛ ∂ρ ⎞ = − , ⎜ ⎟ ρ ′ρ ′′ s′′ − s′ ⎝ ∂ s ⎠p
(3.296)
⎧⎪ C ′′ d ρ ′′ 1 − C ′′ d ρ ′ 1 ⎛ ∂ρ ⎞ ⎡ ds′′ 1 ρ ds′ ⎤ ⎫⎪ = = ρ2 ⎨ 2 + − 2 ⎜ ⎟ ⎢C ′′ + (1 − C ′′ ) ⎥ ⎬ , 2 2 a κp ρ ′ dp ρ ⎝ ∂ s ⎠ p ⎣ dp dp ⎦ ⎭⎪ ⎪⎩ ρ ′′ dp (3.297)
Kolev (1986b). As Eq. (3.296) shows, the density change with the entropy for constant pressure depends strongly on ρ , and therefore on the concentration C”. 50606"Nkswkf/uqnkf"okzvwtg"kp"vjgtoqf{pcoke"gswknkdtkwo" For s′′′ < s < s′, an equilibrium mixture of liquid and solid phases (designated with superscripts ′ and ′′′ respectively) exists. Replacing the superscript ′′ with ′′′ in the formalism of the previous section the required equation of state and the corresponding derivatives are then obtained. 50607"Uqnkf"rjcug" For s < s ′′′ , a solid phase exists. It is assumed here that analytical expressions for the corresponding derivatives are available. Solid phases can be approximately treated as shown in Section 3.4.2.1.
Appendix 3.1 Application of the theory to steam-air mixtures
165
507"Gzvgpukqp"uvcvg"qh"nkswkfu" Superheated liquids are typical examples for liquids in tension state. Their internal pressure is higher than the surrounding pressure and the molecules starts to build clusters of vapor nucleolus. The liquid properties are then computed extrapolating the approximations into the region between the saturation- and the spinoidal line. Liquids at low pressure can be brought in extension state for a very short time too. This state may lead to negative pressures. Many times in the past, researchers have observed in clever experiments such states. Such states are usual if at low pressure strong pressure waves are reflected. Similarly, fast closing valves interrupting fast liquid flows, results also in tension state of the liquid behind the valves. The question of practical interests is haw to compute the properties at negative pressure. The reader will find seldom information to this subject in the literature. Worthington (1892) reported “…In the neighborhood of the zero pressure the absolute coefficient of volume elasticity of alcohol is the same for extension as for compression, and so far as the observation shows is constant between pressure of -17 and atmospheres 12.” It is confirmed also by Meyer (1911), report1 ⎛ ∂v ⎞ ing that “…the extension coefficient of water βT = ⎜ ⎟ was measured dev ⎝ ∂p ⎠T pending on the temperature from 0 to 31°C and pressures between –30 to 7 atmospheres and found in agreement with the compression coefficient at this temperatures.” Skripov et al. (1980) reported properties for heavy water and recommended use of linear change of the density with the pressure in the meta-stable region. Therefore, a useful strategy as a first approximation is to fix the derivatives at the corresponding temperature and minimum pressure and using them to linearly extrapolate densities, enthalpies, entropies etc. Regarding the transport properties, again they can be fixed at the given temperature and minimum pressure for which they are valid and then used in the region of negative pressure until better approach is found.
Crrgpfkz"503"Crrnkecvkqp"qh"vjg"vjgqt{" vq"uvgco/ckt"okzvwtgu" Consider a mixture consisting of steam and air. This mixture is common in nature. To facilitate easy application, the definitions obtained for the partial derivatives are reduced to those for a two-component mixture consisting of one inert and one non-inert component. Here air is designated with A and steam with S, with air taken as a single ideal gas. The result is ⎡ ⎛ ∂ρ ⎞ ⎤ ⎛ ∂ρ ⎞ = 1 ⎢C A RAT + (1 − C A ) / ⎜ S ⎟ ⎥ , ⎜ ⎟ ⎢⎣ ⎝ ∂ p ⎠T ,all _ C ′s ⎝ ∂ pS ⎠T ⎦⎥
(A.1)
166
3 Derivatives for the equations of state
⎡ ⎛ ∂ρ ⎞ ⎛ ∂ρ ⎞ ⎛ ∂ρ S ⎞ ⎛ ∂ρ S ⎞ ⎤ =⎜ ⎢ − ρ A RA + ⎜ ⎟ ⎥, ⎟ ⎜ ⎟ ⎟ ⎜ ⎝ ∂ T ⎠ p ,all _ C′s ⎝ ∂ p ⎠T ,all _ C ′s ⎣⎢ ⎝ ∂ T ⎠pS ⎝ ∂ pS ⎠T ⎥⎦
(A.2)
⎡ ⎤ ⎛ ∂ρ ⎞ ⎛ ∂ρ ⎞ 1 = −ρ ⎜ ⎢ RAT − ⎥, ⎜ ⎟ ⎟ (∂ρ S ∂ pS )T ⎥⎦ ⎝ ∂ p ⎠T ,all _ C′s ⎣⎢ ⎝ ∂ C A ⎠ p ,T ,all _ C′s _ except _ C A
(A.3)
⎧⎪ ( ∂ hS ∂ pS )T ⎡ ⎛ ∂ρ ⎞ ⎛ ∂ρ ⎞ ⎤ ⎫⎪ c p = C A c pA + (1 − C A ) ⎨c pS + − ⎜ S ⎟ ⎥⎬ ⎢(1 − C A ) ⎜ ⎟ ( ∂ρ S ∂ pS )T ⎢⎣ ⎝ ∂ T ⎠ p ,all _ C′s ⎝ ∂ T ⎠ pS ⎦⎥ ⎪⎭ ⎩⎪ = C Ac pA + (1 − C A )c pS ⎡ ( ∂ hS ∂ pS )T ⎛ ∂ρ ⎞ ⎤ − C A (1 − C A ) RA ⎢(1 − C A ) ρ + T ⎜ S ⎟ ⎥ , ⎝ ∂ T ⎠ pS ⎥⎦ 1 − C A ⎡⎣1 − RAT ( ∂ρ S ∂ pS )T ⎤⎦ ⎣⎢ (A.4)
⎛ ∂h ⎞ ⎛∂h⎞ ⎛ ∂ρ ⎞ =⎜ (1 − C A ) 2 ⎜ S ⎟ ⎜ ⎟ ⎟ ⎝ ∂ p ⎠T ,all _ C′s ⎝ ∂ p ⎠T ,all _ C′s ⎝ ∂ pS ⎠T
⎛ ∂ρ S ⎞ ⎜ ⎟ ⎝ ∂ pS ⎠T
⎡ ⎛ ∂ρ S ⎞ ⎤ ⎪⎫ ⎪⎧ ⎨1 − C A ⎢1 − RAT ⎜ ⎟ ⎥⎬, ⎢⎣ ⎝ ∂ pS ⎠T ⎥⎦ ⎪⎭ ⎩⎪
⎛ ∂h ⎞ = (1 − C A ) 2 ⎜ S ⎟ ⎝ ∂ pS ⎠T
(A.5)
⎛ ∂s ⎞ ⎜ ⎟ ⎝ ∂ C A ⎠ p ,T ,all _ C′s _ except _ Ci
= s A − sS −
ρ T
⎡⎛ ∂ hS ⎞ ⎟ ⎢⎣⎝ ∂ pS ⎠T
(1 − CA ) ⎢⎜
⎛ ∂h ⎞ = s A − sS − (1 − C A ) ⎜ S ⎟ ρ RA ⎝ ∂ pS ⎠T
⎤ ⎛ ∂ρ S ⎞ ⎤ ⎡ 1 ⎛ ∂ρ ⎞ ⎥ ⎜ ⎟ ⎥ ⎢1 − (1 − C A ) ⎜ ⎟ ρ ⎝ ∂ C A ⎠ p ,T ,all _ C′s _ except _ Ci ⎥⎦ ⎝ ∂ pS ⎠T ⎥⎦ ⎢⎣ ⎧⎪ ⎡ ⎛ ∂ρ S ⎞ ⎤ ⎫⎪ ⎨1 − C A ⎢1 − RAT ⎜ ⎟ ⎥⎬ , ⎝ ∂ pS ⎠T ⎦⎥ ⎪⎭ ⎪⎩ ⎣⎢
(A.6)
and
⎛∂T ⎞ 1 = ⎜ ⎟ ⎝ ∂ p ⎠ s ,all _ C ′s ρ c p T ⎛ ∂T ⎞ = , ⎜ ⎟ ∂ s c ⎝ ⎠ p ,all _ C′s p
⎡ ⎤ ⎛∂h⎞ ⎢1 − ρ ⎜ ⎥, ⎟ ⎢⎣ ⎝ ∂ p ⎠T ,all _ C ′s ⎥⎦
(A.7)
(A.8)
Appendix 3.2 Useful references for computing properties of single constituents
⎛ ∂T ⎞ T ⎛ ∂s ⎞ =− ⎜ . ⎜ ⎟ ⎟ c p ⎝ ∂ C A ⎠ p ,T ,all _ C′s _ except _ C ⎝ ∂ C A ⎠ p ,s ,all _ C′s _ except _ Ci A
167
(A.9)
Crrgpfkz"504"Wughwn"tghgtgpegu"hqt"eqorwvkpi"rtqrgtvkgu" qh"ukping"eqpuvkvwgpvu"" In this appendix a number of references that provide a useful set of approximations for thermodynamic and transport properties of simple constituents are summarized. These approximations are an example of a simple state equation library. The library discussed below as an example is used in the computer code IVA3 Kolev (1991a, b, c). This library consists of a set of analytical approximations for the following simple substances: air, water, steam, uranium dioxide in solid, liquid and equilibrium solid/liquid state. Alternatively, the analytical approximations for stainless steel or corium can be used instead of the approximation for uranium dioxide properties. For air, the Irvine and Liley (1984) approximations of ρ , c p , h = f (T ) and s = s (T , p = const ) are used where the influence of the pressure on the entropy,
− R ln ( p / p0 ) is added, where p0 is a reference pressure (e.g. 105 ) and R is the
gas constant of air. For steam, the Irvine and Liley (1984) approximations of ρ , c p , s, h = f (T , p ) are used. For water the Rivkin and Kremnevskaya (1977) approximations of ρ , c p , h = f (T , p ) are used. For metastable water, the above analytical properties from Rivkin and Kremnevskaya (1977) are extrapolated, taking into account the discussion by Skripov et al. (1980). The water entropy as a function of temperature and pressure is computed as follows. First the saturation entropy as a function of liquid temperature is computed, s ′ = s′ ⎣⎡ p′ (T ) ⎦⎤ , using analytical approximations proposed by Irvine and Liley (1984), with the pressure correction s = s ⎡⎣ s′, p′ (T ) − p ⎤⎦ then introduced as proposed by Garland and Hand (1989). The analytical approximations by Irvine and Liley (1984) for the steam/water saturation line, p′ = p′ (T ) and T ′ = T ′ ( p ) are used in IVA3. The Clausius-
Clapeyron equation for dp′ / dT was obtained by taking the first derivative of the analytical approximation with respect to temperature. The steam/water saturation properties are computed as a function of p and T ′ ( p ) using the above approximations.
Analytical approximations for the properties ρ , c p , s, h = f (T ) of solid and liquid uranium dioxide as proposed by Fischer (1990), Chawla et al. (1981) and Fink et al. (1982) are used. For the solid stainless steel properties ρ , c p , s, h = f (T ) the analytical approximations proposed by Chawla et al. (1981) are recommended. For liquid
168
3 Derivatives for the equations of state
stainless steel properties ρ , c p , s, h = f (T ) the approximations proposed by
Chawla et al. (1981) and Touloukian and Makita (1970) are recommended. For the solid/liquid two-phase region for liquid metal, the assumptions of thermodynamic equilibrium within the velocity field are used, and the properties then computed as explained in Kolev (1991d). The derivatives ( ∂ h / ∂ p )T , ( ∂ρ / ∂ T ) p , ( ∂ρ / ∂ p )T are easily obtained by differentiating the corresponding analytical approximations. The transport properties of the simple substances are computed as follows. Thermal conductivity and dynamic viscosity of air, and steam λ = λ (T ) - after Irvine and Li-
ley (1984). The water thermal conductivity λ = λ (T , p ) is computed using the Rivkin and Alexandrov (1975) approximation and the water dynamic viscosity η = η (T , p ) and surface tension σ = σ (T ) using the TRAC approximation, Liles et al. (1981). The thermal conductivity and dynamic viscosity of the air-steam mixture are computed using the mole weight method of Wilke (1950). The thermal conductivity of solid and liquid uranium dioxide and stainless steel, as well as the dynamic viscosity and surface tension of liquid uranium dioxide and steel, are computed using the Chawla et al. (1981) approximations. Hill and Miyagawa proposed in 1997 to use The IAPS Industrial Formulation 1997 for the Thermodynamic Properties of Water and Steam, see in Wagner et al. (2000), in tabulated form. The use of an equidistant grid on the two independent axes is proposed and storing not only the properties but also their derivatives at a center of the computational cell. Marking this point with the integer coordinates (i,j) the interpolation is made by Taylor series expansion 2 ⎛ ∂f ⎞ 1 2⎛∂ f ⎞ ⎛ ∂f ⎞ f ( x, y ) = f i , j + ( x − xi ) ⎜ ⎟ + ( x − xi ) ⎜ 2 ⎟ + ( y − yj )⎜ ⎟ ∂ x 2 ∂ x ⎝ ⎠ y ,i , j ⎝ ∂y ⎠ x ,i , j ⎝ ⎠ y ,i , j
+
2 2⎛∂ f ⎞ ⎛ ∂2 f ⎞ 1 y − yj ) ⎜ 2 ⎟ + ( x − xi ) ( y − y j ) ⎜ ( ⎟ 2 ⎝ ∂y ⎠ y ,i , j ⎝ ∂x∂y ⎠i , j
In addition to first derivatives, which are usually available as analytical derivatives from the basic functions, the second derivatives have to be computed numerically, e.g. with steps of 0.001 K and 1000 Pa. Six constants per property and cell center have to be stored. The search of the integer address is quick because of the selected equidistant grid. The reversed functions are also easily obtained by solving analytically the above quadratic equation. Appropriate selected grids contain the critical point at the boundary of a cell (not at the center). Thus, for two such neighboring cells sub- and supercritical approximations are used, respectively. xi and y j may be slightly removed from the center if the saturation line crosses the cell. In this case the properties at the stable site of the saturation line are used. The authors also recommended that instead of the functions v, h, s = f ( p, T ) , the more appropriate pv, h, ps = f ( p, T ) be used, which give finite values for pressures tending to zero.
Appendix 3.3 Useful definitions and relations between thermodynamic quantities
169
Reliable source of water and steam approximations is Wagner and Kruse (2000) providing also some inverted approximations for the sub-critical region. Some inverted approximations for water and steam are also available in Meyer-Pittroff et al. (1969). General approaches for constructing properties for liquids and gases are given in Reid et al. (1982). Large data base for caloric and transport properties for pure substances and mixtures is given in Vargaftik et al. (1996). The NIST-JANAF thermo-chemical tables edited by Chase (1998) are inevitable for computational analysis of reactive flows. Useful differential thermodynamic relationships are given in Appendix 3.3. The reader can find in Elizer et al. (2002) the relativistic modification to this relation.
Appendix 3.3 Useful definitions and relations between thermodynamic quantities ∂ ∂p
v
T s
∂ ∂v
p
T
s
∂ ∂T
p
v
s
∂ ∂s
p
v
T
p 1
v 0
0
s kcv αT −α v
1
−kv
α Tv
0
1
T k
α
−
cp 1 αv 0 −
αT
cp
α Tv α k 0
kcv
1 1 1
cp T cv T 0
0
− −
1 1 kα cp cv vk
1
T cv 0
1
1 1
0
αv
α
0
k cp
α Tv T cp
cv vk cp
0
cv k
αT α Tv cp
αT
0
kcv
1
−
1
k
αv
α
170
3 Derivatives for the equations of state
Specific capacity at constant volume
⎛ ∂u ⎞ cv = ⎜ ⎟ ⎝ ∂T ⎠v
Specific capacity at constant pressure
⎛ ∂u ⎞ ⎛ ∂v ⎞ cp = ⎜ ⎟ + p⎜ ⎟ ⎝ ∂T ⎠ p ⎝ ∂T ⎠ p
Thermal expansion coefficient Compressibility Coefficient of thermal strain
1 ⎛ ∂v ⎞
α= ⎜ ⎟ v ⎝ ∂T ⎠ p 1 ⎛ ∂v ⎞ k =− ⎜ ⎟ v ⎝ ∂p ⎠T
σ=
1 ⎛ ∂p ⎞ ⎜ ⎟ p ⎝ ∂T ⎠v
Tghgtgpegu" Chase MW ed (1998) NIST-JANAF Thermochemical Tables, 4th ed Part I, II, American Institute of Physics and American Chemical Society, Woodbury, New York Chawla TC et al (1981) Thermophysical properties of mixed oxide fuel and stainless steel type 316 for use in transition phase analysis, Nuclear Engineering and Design, vol 67 pp 57-74 Cordfunke EHP, Konings RJM, Eds (1990) Thermo-chemical data for reactor materials and fission products, North-Holland, Amsterdam Elizer S, Ghatak A and Hora H (2002) Fundamental of equation of state, World Scientific, New Jersey Elsner N (1974) Grundlagen der Technischen Thermodynamik, 2. Berichtete Auflage, Akademie-Verlag, Berlin Fink JK, Ghasanov MG, Leibowitz L (May 1982) Properties for reactor safety analysis, ANL-CEN-RSD-82-2 Fischer EA (1990) Kernforschungszentrum Karlsruhe GmbH, unpublished report. Garland WJ, Hand BJ (1989) Simple functions for the fast approximation of light water thermodynamic properties, Nuclear Engineering and Design, vol 113 pp 21-34 Hill PG, Miyagawa K (April 1997) A tabular Taylor series expansion method for fast calculation of steam properties, Transaction of ASME, Journal of Engineering for Gas Turbines and Power, vol 119 pp 485-491 Irvine TF, Liley PE (1984) Steam and gas tables with computer equations, Academic Press, New York Issa RI (1983) Numerical methods for two- and three- dimensional recirculating flows, in Comp. Methods for Turbulent Transonic and Viscous Flow, Ed. J. A. Essers, Hemisphere, Washington, and Springer, Berlin Heidelberg New York, pp 183 Kestin J (1979) A course in thermodynamics, vol 1, Hemisphere, Washington Kolev NI (1986a) Transient three phase three component non-equilibrium nonhomogeneous flow, Nuclear Engineering and Design, vol 91 pp 373-390 Kolev NI (1986b) Transiente Zweiphasenstromung, Springer, Berlin Heidelberg New York
References
171
Kolev NI (Sept. 1991a) A three-field model of transient 3D multi-phase three-component flow for the computer code IVA3, Part 2: Models for interfacial transport phenomena. Code validation, Kernforschungszentrum Karlsruhe, KfK 4949 Kolev NI (Sept. 1991b) A three-field model of transient 3D multi-phase three-component flow for the computer code IVA3, Part 1: Theoretical basics: Conservation and state equations, numerics, Kernforschungszentrum Karlsruhe, KfK 4948 Kolev NI (Sept. 1991c) IVA3: Computer code for modeling of transient three dimensional three phase flow in complicated geometry, Program documentation: Input description, KfK 4950, Kernforschungszentrum Karlsruhe Kolev NI (1991d) Derivatives for equations of state of multi-component mixtures for universal multi-component flow models, Nuclear Science and Engineering, vol 108 no 1 pp 74-87 Kolev NI (1994a) The code IVA4: Modeling of mass conservation in multi-phase multicomponent flows in heterogeneous porous media, Kerntechnik, vol 59 no 4-5 pp 226-237 Kolev NI (1994b) The code IVA4: Modeling of momentum conservation in multi-phase multi-component flows in heterogeneous porous media, Kerntechnik, vol 59 no 6 pp 249-258 Kolev NI (1995) The code IVA4: Second law of thermodynamics for multi phase flows in heterogeneous porous media, Kerntechnik, vol 60 no 1 pp 1-39 Kolev NI (1996) Three fluid modeling with dynamic fragmentation and coalescence fiction or daily practice? 7th FARO Experts Group Meeting Ispra, October 15-16, 1996; Proceedings of OECD/CSNI Workshop on Transient thermal-hydraulic and neutronic codes requirements, Annapolis, MD, U.S.A., 5th-8th November 1996; 4th World Conference on Experimental Heat Transfer, Fluid Mechanics and Thermodynamics, ExHFT 4, Brussels, June 2-6, 1997; ASME Fluids Engineering Conference & Exhibition, The Hyatt Regency Vancouver, Vancouver, British Columbia, CANADA June 22-26, 1997, Invited Paper; Proceedings of 1997 International Seminar on Vapor Explosions and Explosive Eruptions (AMIGO-IMI), May 22-24, Aoba Kinen Kaikan of Tohoku University, Sendai-City, Japan Kolev NI (1998) On the variety of notation of the energy conservation principle for single phase flow, Kerntechnik, vol 63 no 3 pp 145-156 Kolev NI (October 3-8, 1999) Verification of IVA5 computer code for melt-water interaction analysis, Part 1: Single phase flow, Part 2: Two-phase flow, three-phase flow with cold and hot solid spheres, Part 3: Three-phase flow with dynamic fragmentation and coalescence, Part 4: Three-phase flow with dynamic fragmentation and coalescence – alumna experiments, CD Proceedings of the Ninth International Topical Meeting on Nuclear Reactor Thermal Hydraulics (NURETH-9), San Francisco, California Kolev NI (2002) Multi-Phase Flow Dynamics, Vol. 1 Fundamentals + CD, Springer, Berlin, New York, Tokyo, ISBN 3-540-42984-0 Liles DR et al (1981) TRAC-PD2 An advanced best-estimate computer program for pressurized water reactor loss-of-coolant accident analysis, NUREG/CR-2054, LA-7709-MS McCahan S, Shepherd JE (January 1993) A thermodynamic model for aluminum-water interaction, Proc. Of the CSNI Specialists Meeting on Fuel-Coolant Interaction, Santa Barbara, California, NUREC/CP-0127 Meyer J (1911) Negative pressure in liquids, Abh. Dt. Buns. Ges., vol 6 p 1 Meyer-Pittroff R, Vesper H, Grigul U (May 1969) Einige Umkehrfunktionen und Näherungsgleichungen zur “1967 IFC Formulation for Industrial Use” für Wasser und Wasserdampf, Brennst.-Wärme-Kraft, vol 21 no 5 pp 239
172
3 Derivatives for the equations of state
Oelkers EH et al. (1995) Summary of the apparent standard partial molar Gibbs free energies of formation of aqueous species, minerals, and gases at pressures 1 to 5000 bars and temperatures 25 to 1000°C, J. Phys. Chem. Ref. Data, vol 24 no 4 pp 1401-1560 Reid RC, Prausnitz JM, Scherwood TK (1982) The properties of gases and liquids, third edition, McGraw-Hill Book Company, New York Rivkin SL, Alexandrov AA (1975) Thermodynamic properties of water and steam, Energia, Moscow, in Russian Rivkin SL, Kremnevskaya EA (1977) Equations of state of water and steam for computer calculations for process and equipment at power stations, Teploenergetika vol 24 no 3 pp 69-73 (in Russian) Sesternhenn J, Müller B, Thomann H (1999) On the calculation problem in calculating compressible low mach number flows, Journal of Computational Physics, vol 151 pp 579-615 Skripov VP et al (1980) Thermophysical properties of liquids in meta-stable state, Atomisdat, Moscow, in Russian Touloukian YS, Makita T (1970) Specific heat, non-metallic liquids and gases, IFC/Plenum, New York, vol 6 Vargaftik NB, Vonogradov YK and Yargin VS (1996) Handbook of physical properties of liquid and gases, 3d ed., Begell House, New York, Wallingford (UK) Wagner W and Krause A (2000) Properties of water and steam, Springer, Berlin Wagner W et al (2000) The IAPS industrial formulation 1997 for the thermodynamic properties of water and steam, Transaction of ASME, Journal of Engineering for Gas Turbines and Power, vol 122 pp 150-182. See also Wagner W and Kruse A (1998) Properties of water and steam, Springer, Berlin Heidelberg Warnatz J, Maas U, Dibble RW (2001) Combustion, 3th Edition, Springer, Berlin Heidelberg New York Wilke CR (1950) J. Chem. Phys., vol 18 Worthington (1892) Phil. Trans. Royal Soc. London, vol 183 p 355 Julius Meyer (1911) Zur Kenntnis des negativen Druckes in Flüssigkeiten, Verlag von Wilchelm Knapp, Halle a. S.
6"Qp"vjg"xctkgv{"qh"pqvcvkqpu"qh"vjg"gpgti{" eqpugtxcvkqp"hqt"ukping/rjcug"hnqy"
The Greek word τροπη stems from the verb εντροπειη – conversion, transformation, and was used by R. Clausius in a sense of “value representing the sum of all transformations necessary to bring each body or system of bodies to their present state” (1854 Ann. Phys., 125, p. 390; 1868 Phil. Mag., 35, p. 419; 1875 “Die mechanische Wärmetheorie”, vol. 1).
This chapter recalls the achievements of the classical thermodynamics for describing the thermodynamic behavior of flows. At the end of the 19th century this knowledge had already formed the classical technical thermodynamics. The classical thermodynamics makes use of different mathematical notation of the first principles. The purpose of this chapter is to remember that all mathematically correct transformations from one notation into the other are simply logical and consistent reflections of the same physics. The use of the specific entropy is not an exception. The importance of the entropy is that it allows one to reflect nature in the simplest mathematical way. Even being known since 100 years the basic system of partial differential equations is still not analytically solved for the general case. The requirements to solve this system numerically and the virtual possibility to do this by using modern computers is the charming characteristic of the science today. The obtained numerical results have always to be critically examined. As we demonstrate here for pressure wave analysis lower order numerical methods predict acceptably only the first few cycles but then the solution degrades destroyed by numerical diffusion. The inability of the lower order numerical methods to solve fluid mechanics equations for initial and boundary conditions defining strong gradients is not evidence that the equations are wrong. That high accuracy solution methods are necessary to satisfy future needs of the industry is beyond question.
603"Kpvtqfwevkqp" Chapter 4 is intended to serve as an introduction to Chapter 5. The computational fluid mechanics produces a large number of publications in which the mathematical notation of the basic principles and of the thermodynamic relationships is often taken by the authors for self understanding and is rarely explained in detail. One of the prominent examples is the different notation of the energy conservation for flows in the literature. The vector of the dependent
174
4 On the variety of notations of the energy conservation for single-phase flow
variables frequently used may contain either specific energy, or specific enthalpy, or specific entropy, or temperature etc. among other variables. If doing the transformation from one vector into the other correctly the obtained systems of partial differential equation must be completely equivalent to each other. Nevertheless, sometimes the message of the different notations of the first principle is misunderstood by interpreting one of them as “wrong” and other as right. That is why recalling the basics once more seems to be of practical use in order to avoid misunderstandings.
604"Ocuu"cpf"oqogpvwo"eqpugtxcvkqp." gpgti{"eqpugtxcvkqp" The fluid mechanics emerges as a scientific discipline by first introducing the idea of control volume. The flowing continuum can be described mathematically by abstracting a control volume inside the flow and describing what happens in this control volume. If the control volume is stationary to the frame of reference the resulting description of the flow is frequently called Eulerian description. If the control volume is stationary to the coordinate system following the trajectory of fluid particles the description is called Lagrangian. Oswatitsch’s “Gasdynamik” (1952) is my favorite recommendation to start with this topic because of its uncomplicated introduction into the “language of the gas dynamics”. Next we write a system of quasi linear partial differential equations describing the behavior of a single-phase flow in an Eulerian control volume without derivation. The idea of conservation of matter was already expressed in ancient philosophy. Much later, in 1748, M. Lomonosov wrote in a letter to L. Euler “... all changes in the nature happens so that the mass lost by one body is added to other ...”. Equation (4.1) reflects the principle of conservation of mass per unit volume of the flow probably first mathematically expressed in its present form by D’Alembert (1743). The second equation is nothing else than the first principle of Newton (1872) applied on a continuum passing through the control volume probably for the first time by Euler (1737-1755), see Euler (1757). Again the equation is written per unit flow volume. The third equation reflects the principle of the conservation of energy, Mayer (1841-1851), see Mayer (1983), a medical doctor finding that the humans loose more energy in colder regions in the globe then in hotter, which is reflected in the color of the blood due to the different degree of blood oxidation.
∂ρ + ∇ ⋅ ( ρV) = 0 ∂τ ⎛∂V ⎞ + V∇ ⋅ V ⎟ + ∇ ⋅ p + ρ g = 0 ∂τ ⎝ ⎠
ρ⎜
(4.1) (4.2)
4.3 Simple notation of the energy conservation equation 175
∂ ∂τ
⎡ ⎛ 1 2 ⎞⎤ ⎡ ⎛ 1 2⎞ ⎤ ⎢ ρ ⎜ e + 2 V ⎟⎥ + ∇ ⋅ ⎢ ρ ⎜ e + p / ρ + 2 V ⎟ V ⎥ + ρg . V = 0 ⎠⎦ ⎠ ⎦ ⎣ ⎝ ⎣ ⎝
(4.3)
Here the usual notations of the thermodynamics are used. Note that these equations are a)
local volume averaged which means that the behavior of a single molecule interacting with its neighbors is not described in detail.
Secondly, they are b) not time-averaged conservation equation for c) flow without internal or external heat sources. That this equation does not reflect all of the complexity of the flows in nature is evident from the fact that d) the forces resisting the flow due to viscosity are neglected, and consequently e) the power component of these forces into the energy conservation equation is also neglected. Nevertheless the system (4.1) through (4.3) reflects several important characteristics observed in real flows and is frequently used in science and practical computations. In other words this system describes well wave dynamics but does not describe any diffusion processes in the flow, like viscous energy dissipation, heat conduction etc.
605"Ukorng"pqvcvkqp"qh"vjg"gpgti{"eqpugtxcvkqp"gswcvkqp" The energy conservation equation can be substantially simplified as it will be demonstrated here without any loss of generality. To achieve this first the scalar product of Eq. (4.2) is constructed with the field velocity V. The result is a scalar expressing the mechanical energy balance. Subtracting this result from the energy equation results in
∂ ( ρ e ) + ∇ ⋅ ( ρ eV ) + p∇ ⋅ V = 0. ∂τ
(4.4)
Obviously, this equation is much more compact than the primitive form of the energy conservation equation (4.3). Additional simplification is reached if differentiating the first two terms and comparing with the mass conservation equation. Dividing by the density results in
∂e p + ( V ⋅ ∇ ) e + ∇ ⋅ V = 0. ∂τ ρ
(4.4)
Note, that the system consisting of Eqs. (4.1, 4.2, 4.3) is completely equivalent to the system consisting of Eqs. (4.1, 4.2, 4.4 or 4.4a).
176
4 On the variety of notations of the energy conservation for single-phase flow
606"Vjg"gpvtqr{" Equation (4.4) is called frequently in the literature also energy equation. It can be additionally simplified by introducing a new variable. In a closed system the introduced external heat, dq, may be transformed into internal specific energy de, or into expansion work pdv, or into both simultaneously, in a way that no energy is lost dq = de + pdv – energy conservation principle. A unique functional dependence of the type q = q(e,v) cannot be derived. In the mathematics the linear form dq = de + pdv in which dq is not a total differential is called Pfaff´s form. The Gibbs relation, Tds = de + pdv = de −
p
ρ2
dρ,
(4.5)
Gibbs (1878) simply says that per unit mass the change of the internal energy de and the expansion work pdv in the control volume, can be expressed as a change of an single variable denoted with s, provided T is the so-called integrating denominator. That T is the integrating denominator was proven 30 years earlier. The integrating denominator itself was found to be the absolute temperature T (W. Thomson 1848). M. Plank showed that among the many choices for the integrating denominator only the absolute temperature allows to integrate independently from the trajectory of the transformation Plank (1964). The new remarkable variable was called by R. Clausius specific entropy, see Clausius (1854). Specific – because related to unit mass of the fluid. The Greek word for entropy, τροπη , stems from the verb εντροπειη which means conversion or transformation, and is used by Clausius in a sense of “...value representing the sum of all transformations necessary to bring each body or system of bodies to their present state”. The introduction of the specific entropy allows simple notation of the second principle of the thermodynamics formulated also by Clausius. Gibbs used Eq. (4.5) as a mathematical expression of the so called equilibrium principle “... after the entropy of the system reaches the maximum the system is in a state of equilibrium ...”. Equation (4.5) has another remarkable property. It is the differential expression of a unique equation of state s = s ( e, v ) in which the partial derivatives ( ∂ e ∂ s )v = T and ( ∂ e ∂ v ) s = − p are technically measurable properties. Note that the introduction of the new variable entropy plays also an important role in the statistical mechanics and in the informatics which will not be discussed here. Thus, making use of the Gibbs relation, of the mass conservation equation, and of Eq. (4.5), Eq. (4.4a) is easily transformed into the simplest notation expressing the energy conservation law
∂s + ( V ⋅ ∇ ) ⋅ s = 0. ∂τ
(4.6)
4.6 Variety of notation of the energy conservation principle 177
Again note that the system consisting of Eqs. (4.1, 4.2, 4.6) is completely equivalent to the system consisting of Eqs. (4.1, 4.2, 4.4), and consequently to the original system containing Eqs. (4.1, 4.2, 4.3).
607"Gswcvkqp"qh"uvcvg" The macroscopic behavior of fluids can be described by describing the behavior of the simple molecule in interaction with others in a control volume Boltzman (1909). The technical thermodynamics makes use of the so called equation of state in which the local volume- and time-averaged behavior of the fluid is reflected rather than describing the behavior of the single molecule (see the Riemann´s work from 1859 in which he combines the experimental findings by Gay-Lussac (1806), Boyle-Mariotte to one equation of state, or any thermodynamic text book). One example for the equation of state is the dependence of the specific internal energy on the volume-averaged density and on the volume-averaged temperatures, the so called caloric equation of state, e = e( ρ , T ) .
(4.7)
Note the useful definition of the specific capacity at constant volume ⎛ ∂e ⎞ ⎛ ∂e ⎞ ⎜ ⎟ =⎜ ⎟ = cv ∂ T ⎝ ⎠ ρ ⎝ ∂ T ⎠v
(4.8)
Perfect fluids are defined by the dependence of the specific internal energy on the temperature only, but not on the density (remember the famous expansion experiments by Gay-Lussac). Variety of forms of the equation of state is possible which may lead to a variety of notations of the energy conservation principle.
608"Xctkgv{"qh"pqvcvkqp"qh"vjg"gpgti{" eqpugtxcvkqp" rtkpekrng" 60803"Vgorgtcvwtg"
Equation (4.4) can be rewritten in terms of temperature using the equation of state (4.7) and the mass conservation equation (4.1). The result is
∂T p ⎡ ρ2 ⎛ ∂ e ⎞ ⎤ + (V⋅ ∇)T + ⎢1 − ⎜ ⎟ ⎥ ∇ ⋅ V = 0, ∂τ ρ cv ⎣ p ⎝ ∂ρ ⎠T ⎦ an equation containing only one time derivative. Note that
(4.9)
178
4 On the variety of notations of the energy conservation for single-phase flow
p ⎡ ρ2 ⎛ ∂ e ⎞ ⎤ ⎛ ∂ p ⎞ ⎢1 − ⎜ ⎟ ⎥ ≡⎜ ⎟, T⎣ p ⎝ ∂ρ ⎠T ⎦ ⎝ ∂ T ⎠ v
(4.10)
which is an important thermodynamic relationship, see Reynolds and Perkins (1977) p. 266 Eq. 8.54. Again note that the system consisting of Eqs. (4.1, 4.2, 4.9) is completely equivalent to the system consisting of Eqs. (4.1, 4.2, 4.4a), and consequently to the original system containing Eqs. (4.1, 4.2, 4.3). 60804"Urgekhke"gpvjcnr{"
Another specific variable frequently used in the thermodynamic is the specific enthalpy defined as follows h = e + p / ρ.
(4.11)
Using the differential form of this definition d ( ρ e ) + dp = d ( ρ h )
(4.12)
Eq. (4.4) can be written in terms of the specific enthalpy and pressure
∂ ∂p ( ρ h ) + ∇ ⋅ ( ρ hV ) − ⎛⎜ + V ⋅ ∇p ⎞⎟ = 0 . ∂τ ⎝ ∂τ ⎠
(4.13)
Again note that the system consisting of Eqs. (4.1, 4.2, 4.13) is completely equivalent to the system consisting of Eqs. (4.1, 4.2, 4.4), and consequently to the original system containing Eqs. (4.1, 4.2, 4.3). Moreover, it is easily shown that using the definition equation for the specific enthalpy (4.11) the Gibbs relation (4.5) takes the form T ρ ds = ρ dh − dp
(4.14)
which immediately transforms Eq. (4.13) into the entropy equation (4.6) again. Equation (4.13) is easily transferred in terms of temperature and pressure by using the following equation of state h = h(T , p )
(4.15)
or in differential form ⎛∂h⎞ dh = c p dT + ⎜ ⎟ dp. ⎝ ∂ p ⎠T
(4.16)
The resulting equation is ⎛ ∂ h ⎞ ⎤ ⎡∂ p ∂T 1 ⎡ ⎤ + (V ⋅ ∇)T − + ( V ⋅ ∇ ) p ⎥ = 0. ⎢1 − ρ ⎜ ⎟ ⎥⎢ ∂τ ρcp ⎣ ⎦ ⎝ ∂ p ⎠T ⎦ ⎣ ∂τ
(4.17)
4.8 The equivalence of the canonical forms 179
Again note that the system consisting of Eqs. (4.1, 4.2, 4.17) is completely equivalent to the system consisting of Eqs. (4.1, 4.2, 4.13), and consequently to the original system containing Eqs. (4.1, 4.2, 4.3).
609"Uwooct{"qh"fkhhgtgpv"pqvcvkqpu" It follows from Sections 4.3 through 4.6 that the system of equations (4.1, 4.2, 4.3) is completely equivalent to one of the systems containing equations (4.1) and (4.2), and one of the following expressions of the first principle
∂e p + ( V ⋅ ∇ ) e + ∇ ⋅ V = 0, ∂τ ρ ∂s + ( V ⋅ ∇ ) s = 0, ∂τ
∂T p + (V ⋅ ∇)T + ∂τ ρ cv
(4.4a) (4.6)
⎡ ρ2 ⎛ ∂ e ⎞ ⎤ ⎢1 − ⎜ ⎟ ⎥ ∇ ⋅ V = 0, p ⎝ ∂ρ ⎠T ⎦ ⎣
(4.9)
∂h 1 ⎡∂ p ⎤ + (V ⋅ ∇) h − ⎢ + ( V ⋅ ∇ ) p ⎥ = 0, ∂τ ρ ⎣ ∂τ ⎦
(4.13)
⎛ ∂ h ⎞ ⎤ ⎡∂ p ∂T 1 ⎡ ⎤ + ( V ⋅ ∇)T − + ( V ⋅ ∇ ) p ⎥ = 0. ⎢1 − ρ ⎜ ⎟ ⎥⎢ ∂τ ρcp ⎣ ⎦ ⎝ ∂ p ⎠T ⎦ ⎣ ∂τ
(4.17)
These equations are derived without specifying the type of the fluid as perfect or real gas. They are valid for both cases. The reader will find in McDonald et al. (1978) notation using as dependent variables: (w, h, p), (G, h, p), (w, ρ , p), (G, ρ , p) , (w, ρ , h), (G, ρ , h), (w, ρ , e), (G, ρ , e) and in Thorley and Tiley (1987) ( ρ , h, w) and (p, T, w).
60:"Vjg"gswkxcngpeg"qh"vjg"ecpqpkecn"hqtou" The system of partial differential equations can be further considerably simplified if they are from a special type called hyperbolic. Riemann found that such a system can be transferred to the so called canonical form which consists of ordinary differential equations. Next we will start with two different forms of the system, one using specific entropy and the other using specific internal energy. We then give the eigen values, the eigen vectors and the canonical form of both systems. It will be shown again that the so obtained forms are mathematically identical.
180
4 On the variety of notations of the energy conservation for single-phase flow
Before performing the analysis of the type of the system we simplify the mass conservation equation as follows. The density in the mass conservation equation is expressed as function of pressure and specific entropy
ρ = ρ ( p, s )
(4.18)
or in differential form dρ =
dp ⎛ ∂ρ ⎞ + ⎜ ⎟ ds. a2 ⎝ ∂ s ⎠ p
(4.19)
Substituting Eq. (4.19) into Eq. (4.1) and taking into account the entropy Eq. (4.6) the mass conservation equation takes the form of Eq. (4.21). Thus using the following vector of dependent variables U = ( p, w, s )
(4.20)
the flow is completely described by the system of quasi linear partial differential equations
∂p ∂p ∂w +w + ρ a2 = 0, ∂τ ∂z ∂z
(4.21)
∂w ∂w 1 ∂ p +w + + g z = 0, ∂τ ∂z ρ ∂z
(4.2)
∂s ∂s +w = 0. ∂τ ∂z
(4.6)
It is well known that the eigen values of this system defined by w−λ 1/ ρ 0
ρ a2 w−λ 0
0 0 = 0, w−λ
(4.22)
are real and different from each other
λ1,2 = w ± a , λ3 = w.
(4.23, 4.24, 4.25)
The eigen vectors of the transposed characteristic matrix ⎛ w − λi ⎜ 2 ⎜ ρa ⎜ 0 ⎝
1/ ρ w − λi 0
⎞ ⎛ h1 ⎞ ⎟⎜ ⎟ 0 ⎟ ⎜ h2 ⎟ = 0 , w − λi ⎟⎠ ⎜⎝ h3 ⎟⎠i 0
are linearly independent
(4.26)
4.8 The equivalence of the canonical forms 181
⎛1/ ( ρ a ) ⎞ ⎛ −1/ ( ρ a ) ⎞ ⎛0⎞ ⎜ ⎟ ⎜ ⎟ ⎜ ⎟ h1 = ⎜ 1 ⎟ , h2 = ⎜ 1 ⎟ , and h3 = ⎜ 0 ⎟. ⎜1⎟ ⎜ 0 ⎟ ⎜ ⎟ 0 ⎝ ⎠ ⎝ ⎠ ⎝ ⎠
(4.27, 4.28, 4.29)
Therefore the quasi linear system of partial differential equations is hyperbolic and can be transformed into the very simple and useful canonical form dz = w+ a, dτ dz = w−a , dτ dz = w, dτ
1 dp
ρ a dτ −
+
dw = −g, dτ
1 dp dw + = −g, ρ a dτ dτ
ds = 0. dτ
(4.30, 4.31) (4.32, 4.33) (4.34, 4.35)
As far as I know Eqs. (4.30 to 4.35) were obtained for the first time by Riemann in 1859. Remember that each of the ordinary differential equations (4.31, 4.33, 4.35) is valid in its own coordinate system attached along one of the three characteristic lines defined by the equations (4.30, 4.32, 4.34), respectively. Equations (4.34, 4.35) contain a very interesting analytical solution s = const. It means that an infinitesimal control volume moving with the fluid velocity along the line defined by Eq. (4.34) in the (τ , z ) plane does not change a specific property defined as specific entropy. This change of state is called isentropic. Comparing with Eq. (4.5) it simply means that within this control volume the energy required for volume expansion per unit flow volume reduces the specific internal energy per unit time by the same amount. Now let us perform the same analysis starting from the system containing the first principle in terms of specific internal energy. We use this time as dependent variables vector, U = ( p, w, e ) ,
(4.36)
and the system describing the flow
∂p ∂p ∂w +w + ρa2 =0 ∂τ ∂z ∂z
(4.21)
∂w ∂w 1 ∂ p +w + + gz = 0 ∂τ ∂z ρ ∂z
(4.2)
∂e ∂e p ∂w +w + = 0. ∂τ ∂z ρ ∂z
(4.4a)
If you are not familiar with the analysis of the type of a system of partial differential equations by first computing the eigen values, eigenvectors and canonical forms it is recommended first to read Section 11 before continuing here.
182
4 On the variety of notations of the energy conservation for single-phase flow
From the characteristic equation w−λ 1/ ρ
ρ a2 w−λ
0 0
p
w−λ
0
ρ
=0
(4.37)
we obtain the same eigen values, Eqs. (4.23, 4.24, 4.25), as in the previous case, which is not surprising for equivalent systems. Using the eigen vectors of the transposed characteristic matrix ⎛w−λ ⎜ ⎜ ρ a2 ⎜ ⎜⎜ ⎝ 0
1/ ρ w−λ 0
0 ⎞ ⎟ ⎛ h1 ⎞ p ⎟⎜ ⎟ h2 = 0 , ρ ⎟ ⎜⎜ ⎟⎟ ⎟ h3 w − λ ⎟⎠ ⎝ ⎠
(4.38)
we can transform the system of partial differential equations again into a system of ordinary equations along the characteristic lines. Along the first two characteristic lines both canonical equations are identical with the system using the specific entropy as a dependent variable. Along the third characteristic line we have a canonical equation which is in terms of specific internal energy and pressure, dz = w, dτ
2
de ⎛ p ⎞ 1 dp −⎜ = 0. ⎟ dτ ⎝ ρ a ⎠ p dτ
(4.39, 4.40)
For the general case it can be shown that the following relation holds 2
⎡ ⎛ p ⎞ 1 p ⎛ ∂ρ ⎞ ⎤ dp . ⎢T + 2 ⎜ ⎟ ⎥ ds = de − ⎜ ⎟ ρ ⎝ ∂ s ⎠ p ⎥⎦ ⎢⎣ ⎝ ρa ⎠ p
(4.41)
Applying this relation to the canonical equation (4.40) results exactly in Eq. (4.35). This demonstrates that also the canonical forms of the system using different sets as depending variables are mathematically equivalent.
60;"Gswkxcngpeg"qh"vjg"cpcn{vkecn"uqnwvkqpu" That the analytical solutions of the equivalent systems of equations written in terms of different dependent variables are equivalent to each other is obvious.
4.10 Equivalence of the numerical solutions? 183
6032"Gswkxcngpeg"qh"vjg"pwogtkecn"uqnwvkqpuA" 603203"Gzrnkekv"hktuv"qtfgt"ogvjqf"qh"ejctcevgtkuvkeu"
Next let us apply the simple explicit first order method of characteristics for solving the system for a horizontal equidistant discretized pipe, Δz1 = ( w + a ) Δτ ,
1
ρa
(p
τ +Δτ
− p1 ) + wτ +Δτ − w1 = 0,
1 ( pτ +Δτ − p2 ) + wτ +Δτ − w2 = 0 , ρa
Δz2 = ( w − a ) Δτ ,
−
Δz3 = wΔτ ,
sτ +Δτ − s3 = 0 ,
(4.42, 4.43) (4.44, 4.45) (4.46, 4.47)
where indices i refers to values of the dependent variables at (τ , z − λi Δτ ) . There are a variety of methods to compute U i as a function of the neighboring values. One of the simplest is the linear interpolation Ui = +
1 ⎡1 + sign ( Cui )⎤⎦ ⎡⎣U k −1 + (U − U k −1 )(1 − Cui )⎤⎦ 2⎣
1 ⎡1 − sign ( Cui )⎤⎦ ⎡⎣U − (U k +1 − U ) Cui ⎤⎦ , 2⎣
(4.48)
where Cui =
λi Δτ Δz
.
(4.49)
The result is wτ +Δτ =
⎤ 1⎡ 1 − ( p2 − p1 ) + w1 + w2 ⎥ ⎢ 2 ⎣ ρa ⎦
(4.50)
pτ +Δτ =
1 ⎡ p2 + p1 − ( w2 − w1 ) ρ a ⎤⎦ 2⎣
(4.51)
sτ +Δτ = s3 .
(4.52)
Obviously, the pressure and the velocity in the point (τ + Δτ , z ) depend on the pressure and the velocities at points 1 and 2. The old time value of the specific entropy is simply shifted along the third characteristic from the point (τ , z − wΔτ ) to the point (τ + Δτ , z ) . This result is derived without any assumption of the type of the fluid.
184
4 On the variety of notations of the energy conservation for single-phase flow
In case of closed ends of the pipe we have the following results. For the left end wL = 0 , s3 = sL , and therefore sτ +Δτ = sL , and from Eq. (4.45) pτ +Δτ = p2 − w2 ρ a . Similarly for the right end wR = 0 , s3 = sR , and therefore
sτ +Δτ = sR , and from Eq. (4.43) pτ +Δτ = p1 + w1 ρ a . For open ends only the canonical equations whose characteristics are inside the pipe are used. For the missing equations boundary conditions have to be provided. Joukowsky (1898) derived the closed end relation already in 1898. He explained for the first time the physics behind the maximum pressure spikes for water pipe lines in which valves are closed quickly. Therefore, fast closing valves produce stagnation pressure spikes that are proportional (a) to the undisturbed initial velocity and (b) to the sound velocity, pτ + Δτ − p1 = ρ w1a , Eq. (16), Joukowsky (1898).
This is important knowledge for designing the pipe supports in the engineering practice. Either the designer of the valves have to implement solution not allowing fast closing, or if fast closing is desired for some reasons, the pipes have to sustain the forces originating after the closing. Farther more Joukowsky uses the Riemann’s method to transfer the mass and momentum equations into canonical forms and proposed for the first time numerical solution method which is known now as the “method of characteristics”. Not having computers, Joukowsky proposed graphical method of characteristics. In his important for the modern hydrodynamics work, Joukowsky also considered the influence of the elasticity of the pipe walls on the velocity of sound, Eq. (15) in Joukowsky (1898). The reader will find examples for numerical solutions based on the singlephase method of characteristics as follows: 1D-flows: For diabatic three-equations model see Namatame and Kobayashi (1974), for two-equations model for pipe-networks see Katkovskiy and Poletaev (1975), Henshaw (1987), Weisman and Tenter (1981), for two-equations model for pipe-networks with pumps see Capozza (1986), Chen et al. (1974), Shin and Chen (1975). 2D-flows: For single component flow see George and Doshi (1994), Henshaw (1987), Weisman and Tenter (1981), for multi-component steady reactive flows see Zucrow and Hoffman (1977), for isentropic flow see Shin and Valentin (1974). 1D homogeneous two-phase flows: The extension of the single phase flow models to two-phase equilibrium models is very easy. The mixture specific entropy is defined as follows
4.10 Equivalence of the numerical solutions? 185
s = X 1,eq s′′ + (1 − X 1,eq ) s ′, resulting in equilibrium vapor mass flow fraction X 1, eq =
s − s′ . s ′′ − s ′
The specific mixture volume is then v = X 1,eq v′′ + (1 − X 1,eq ) v ′ =s
v′′ − v′ s ′′v′ − s ′v′′ dT ′ s ′′v′ − s ′v′′ + =s + = f ( s, p ) = sf1 + f 2 = v ( p, s ). s ′′ − s ′ s′′ − s′ dp s′′ − s′
Therefore a convenient equation of state is v = v ( p, s ) with differential form ⎛ ∂v ⎞ ⎛ ∂v ⎞ dv = ⎜ ⎟ dp + ⎜ ⎟ ds, ∂ p ⎝ ∂s ⎠ p ⎝ ⎠s
where ⎛ ∂v ⎞ df1 df 2 + , ⎜ ⎟ =s ∂ p dp dp ⎝ ⎠s
⎛ ∂v ⎞ ⎜ ⎟ = f1 . ⎝ ∂s ⎠ p We realize that the partial derivative of the mixture specific volume with respect to the specific mixture entropy is equivalent to the Clapeyron’s expression dT ′ v′′ − v′ = . dp s ′′ − s ′
Approximating the function f1 and f2 with smooth pressure functions is very convenient for using them in fast computational algorithms. The velocity of sound a is easily derived if we use the above expressions to obtain the equation of state ρ = 1 v = ρ ( p, s ) or in differential form dρ =
dp ⎛ ∂ρ ⎞ + ⎜ ⎟ ds , a 2 ⎝ ∂s ⎠ p
where 1 ⎛ ∂ρ ⎞ 1 ⎛ df1 df 2 ⎞ =⎜ + ⎟ = − 2 ⎜s ⎟ 2 a v ⎝ dp dp ⎠ ⎝ ∂p ⎠ s
186
4 On the variety of notations of the energy conservation for single-phase flow
⎛ ∂ρ ⎞ 2 ⎜ ⎟ = − f1 v . ⎝ ∂s ⎠ p Alternative expression for the velocity of sound, 1
( ρa)
2
=
v′′ − v′ ⎡ ds′′ ds′ ⎤ dv′′ dv′ + (1 − X 1, eq ) ⎥ − X 1, eq + (1 − X 1, eq ) , ⎢ X 1, eq s ′′ − s′ ⎣ dp dp ⎦ dp dp
is frequently used in the literature but it is obviously computationally more expensive compared to the previous expression. Homogeneous non equilibrium 4-equation model is solved by the method of characteristics on structured grid using second order method by Köberlein (1972), Grillenberger (1981). Homogeneous equilibrium 3-equations models with variety of notations for the energy conservation are reported by McDonald et al. (1978). Different variants of the methods of characteristics are proposed in this work with the most interesting one based on two grids. On the even grids a conservative form of the conservation equations is used guaranteeing strict conservation, and on the odd grids – the canonical form using the method of characteristics. Homogeneous equilibrium 3-equations and homogeneous nonequilibrium 5-equations model are solved by the method of characteristics on non structured grid by Hancox and Banerjee (1977), Hancox et al. (1978). Homogeneous non-equilibrium 5-equations model is solved by the method of characteristics on non structured grid by Ferch (1979); For two fluid model without thermal interaction see Weisman and Tenter (1981). 1D non-homogeneous two-phase flows: For isentropic drift flux model see Kaizerman et al. (1983). Analytical solution for concentration waves in boiling channel at constant pressure and known mass flow is provided by Santalo and Lahey (1972). If the spatial derivatives in the canonical form are computed by finite differences for the old time plane by using weighing depending on the direction of the velocity the method is called method of lines or pseudo-characteristics method. Examples are available by Wang and Johnson (1981), Kolev and Carver (1981). Positioning adaptively more grid points along the pipe axis where the gradients of the variables are stronger is a efficient technique to increase the accuracy of the method of lines by low computational costs Hu and Schiesser (1981). To get an impression about the physical meaning of Eq. (4.52) we rewrite it for a perfect gas T τ +Δτ ⎛ pτ +Δτ ⎞ =⎜ ⎟ T3 ⎝ p3 ⎠
R / cp
.
(4.53)
This is usually used as a benchmark for any other numerical solution using first order explicit methods.
4.10 Equivalence of the numerical solutions? 187
Applying again the explicit method of characteristics to the system in terms of the specific internal energy we obtain the same results for the pressure and velocities. The relation along the third characteristic line should be considered very carefully. If the differential form 2
⎛ p ⎞ dp de = ⎜ ⎟ ⎝ ρa ⎠ p
(4.54)
is analytically integrated for a perfect gas it results exactly in Eq. (4.53). This is not the case if the differential form is written in finite difference form, eτ +Δτ = e3 +
p
( ρa)
(p
τ +Δτ
2
− p3 ) ,
because for the computation of the coefficient
(4.55) p
( ρa)
2
an assumption has to be
made of how it has to be computed. For this reason the result, T τ +Δτ − T3 =
T R τ +Δτ ( p − p3 ) , p cp
(4.56)
does not fit to the exact benchmark solution. Over many time steps it may accumulate to considerable error. 603204"Vjg"rgthgev"icu"ujqem"vwdg0
u≤0 rh ri +1
u *i +1 = u
u* = ui −1
u *i +1 = ui +1
rh ,i −1
u* = u
r
rh ,i +1 ri +1
rh r
Replacing in the Bernulli equation results in
ρ
r ⎞⎛ r r ⎞ 1 1 1 ⎛ rh + ui −1 h, i −1 ⎟ ⎜ u h − ui −1 h, i −1 ⎟ + ( pi +1 − p ) = 0 for u > 0 ⎜u Δrh 2 ⎝ ri +1 r ⎠ ⎝ ri +1 r ⎠ Δrh
and
ρ
rh, i +1 r r ⎞⎛ r 1 1⎛ + u h ⎟ ⎜ ui +1 h, i +1 − u h ⎜ ui +1 Δrh 2 ⎝ ri +1 r ⎠⎝ ri +1 r
⎞ 1 ( pi +1 − p ) = 0 for ⎟+ ⎠ Δrh
(
u ≤ 0,
)(
respectively. We multiply and divide the convective term by u + ui −1 u − ui −1 and use outside this product ui −1 = u
rh rh ,i −1
.
The result is 1 1 − 1 ri 2+1 r 2 1 1 ρ u + ui −1 u − ui −1 + ( pi +1 − p ) = 0 , 1 1 Δrh 2 Δrh − rh2 rh2,i −1
(
)(
)
For the singularity rh ,i −1 = 0 we have
ρ
1 Δrh
2
⎛ rh ⎞ 1 1 u + ui −1 u − ui −1 + ( pi +1 − p ) = 0 . ⎜ ⎟ Δrh ⎝ ri +1 ⎠ 2
(
)(
)
Similarly for negative velocity between the two pressures we have ui +1 = u
rh rh ,i +1
u > 0.
)
572
12 Numerical solution methods for multi-phase flow problems
and therefore 1 1 − 1 ri2+1 r 2 1 1 ρ ( ui +1 + u )(ui +1 − u ) + ( pi +1 − p ) = 0 , 1 1 Δrh 2 Δrh − rh2,i +1 rh2
u ≤0.
Thus the final form of the multiphase convective term is written as 1 1 − 2 2 1 ri +1 r ⎧ 1 ⎫ − ⎨− min ⎡⎣0, (α l ρl ul γ r )1 ⎤⎦ + (α l ρ lν lγ r )1 ⎬ ( uli +1 − ul ) 1 1 Δrh Δri +1 ⎭ ⎩ − rh2 rh2,i −1 1 1 − 1 ri 2+1 r 2 ⎧ 1 ⎫ − ⎨max ⎡⎣ 0, (α l ρl ul γ r )2 ⎤⎦ + (α l ρ lν l γ r )2 ⎬ ( uli −1 − ul ) . 1 1 Δrh Δr ⎭ ⎩ − rh2,i +1 rh2
It is convenient to compute the geometry coefficients at the start
β r1 =
1 1 1 , β r 2 = β r1 , β r 3 = , βr 4 = βr3 , βr5 = , βr6 = βr5 . κ Δrh Δ z ( r + Δrh / 2 ) Δθ
The bu coefficients are defined as follows:
bul1 = − β r1
1 1 − ri 2+1 r 2 1 1 − rh2 rh2,i −1
⎧ 1 ⎫ ⎨− min ⎡⎣0, (α l ρl ul γ r )1 ⎤⎦ + (α l ρ lν l γ r )1 ⎬, Δri +1 ⎭ ⎩
bul 2 = − β r 2
1 1 − ri 2+1 r 2 1 1 − rh2,i +1 rh2
⎧ 1 ⎫ ⎨max ⎡⎣0, (α l ρl ul γ r ) 2 ⎤⎦ + (α l ρlν l γ r ) 2 ⎬, Δr ⎭ ⎩
⎧ 1 ⎫ bul 3 = − β r 3 ⎨− min ⎡⎣0, (α l ρl vl γ θ )3 ⎤⎦ + (α l ρlν l γ θ )3 κ ⎬, r Δ θh ⎭ h ⎩ ⎧⎪ ⎫⎪ 1 bul 4 = − β r 4 ⎨max ⎡⎣0, (α l ρl vl γ θ )4 ⎤⎦ + (α l ρlν l γ θ )4 κ ⎬, rh Δθ h , j −1 ⎪⎭ ⎪⎩
12.17 Pipe networks: some basic definitions
573
⎧ 1 ⎫ bul 5 = − β r 5 ⎨− min ⎡⎣0, (α l ρl wl γ z )5 ⎤⎦ + (α l ρlν l γ z )5 ⎬, Δz h ⎭ ⎩ 1 ⎪⎫ ⎪⎧ bul 6 = − β r 6 ⎨ max ⎡⎣0, (α l ρl wl γ z )6 ⎤⎦ + (α l ρlν l γ z )6 ⎬. Δzh , k −1 ⎭⎪ ⎩⎪
Now we introduce the volume weighting coefficient C* =
Vi +1 . V + Vi +1
This coefficient allows us easily to compute the mass flows entering the staggered cell through each face divided by the staggered cell volume by using the Blm − flux densities already computed for the non-staggered cells ⎛ Du bul ,m = −C * ⎜ Blm − + β m l , m ⎜ ΔLh , m ⎝
⎞ * ⎟⎟ − 1 − C ⎠
(
)
⎛ Dlu, m ⎜⎜ Blm − + β m ΔLh , m ⎝
⎞ ⎟⎟ . ⎠i +1
With this abbreviation we have 6
γ vuα lua ρlua ( ul − ula ) / Δτ + ∑ bul , m ( ul , m − ul ) ... m =1
Crrgpfkz"3407"Vjg" a "eqghhkekgpvu"hqt"Gs0"*34068+" a11 = a22 a33 − a32 a23
a12 = a32 a13 − a12 a33
a13 = a12 a23 − a22 a13
a21 = a23 a31 − a21a33
a22 = a11a33 − a31 a13
a23 = a21a13 − a23 a11
a31 = a21a32 − a31a22
a32 = a12 a31 − a32 a11
a33 = a11a22 − a21a12
Crrgpfkz"3408"Fkuetgvk|cvkqp"qh"vjg"cpiwnct"oqogpvwo"gswcvkqp" The θ momentum equation discretized by using the donor cell concept is
γ vvα lva ρlva ( vl − vla ) / Δτ −
rhκ ⎧ 1 ⎫ ⎨− min ⎡⎣ 0, (α l ρl ul γ r )1 ⎤⎦ + (α l ρlν l γ r )1 ⎬ ( vli +1 − vl ) κ r Δr ⎩ Δrh ⎭
574
12 Numerical solution methods for multi-phase flow problems
−
rhκ,i −1 ⎧⎪ 1 ⎫⎪ ⎨max ⎡⎣ 0, (α l ρl ul γ r )2 ⎤⎦ + (α l ρlν l γ r )2 ⎬ ( vli −1 − vl ) κ r Δr ⎩⎪ Δrh ,i −1 ⎭⎪
−
1 r Δθ h
−
1 ⎧ 1 ⎫ ⎨max ⎡⎣ 0, (α l ρl vl γ θ )4 ⎤⎦ + (α l ρ lν l γ θ )4 κ ⎬ ( vlj −1 − vl ) r Δθ h ⎩ r Δθ ⎭
−
1 ⎧ 1 ⎫ ⎨ − min ⎡⎣0, (α l ρl wl γ z )5 ⎤⎦ + (α l ρlν l γ z )5 ⎬ ( vlk +1 − vl ) Δz ⎩ Δz h ⎭
−
1 ⎧⎪ 1 ⎫⎪ ⎨ max ⎡⎣0, (α l ρl wl γ z )6 ⎤⎦ + (α l ρlν l γ z )6 ⎬ ( vlk −1 − vl ) Δz ⎩⎪ Δzh , k −1 ⎭⎪
+
κ
⎧⎪ 1 ⎫⎪ ⎨− min ⎡⎣0, (α l ρl vl γ θ )3 ⎤⎦ + (α l ρlν l γ θ )3 κ ⎬ ( vlj +1 − vl ) r Δθ j +1 ⎭⎪ ⎩⎪
κ
sin ( Δθ / 2 ) r κ Δθ h
⎧ ⎧ ⎡ ∂ ⎛ vl ⎪ ⎪(αρν )lj +1 γ θ 3 ⎢ r ⎜ ⎪ ⎣ ∂r⎝ r ⎪ ⎪ ⎪ × ⎨(α l ρl vl γ θ )3 ul 3 + (α l ρl vl γ θ )4 ul 4 − ⎨ ⎪ ⎪ ⎪ ⎪ + (αρν ) γ ⎡ r ∂ ⎛ vl ⎜ l θ4 ⎢ ⎪ ⎪ ⎣ ∂r⎝ r ⎩ ⎩ +α lva
⎞ 1 ∂ ul ⎤ ⎫⎫ ⎟ + r ∂θ ⎥ ⎪⎪ ⎠ ⎦ 3 ⎪⎪ ⎪⎪ ⎬⎬ ⎪⎪ ⎞ 1 ∂ ul ⎤ ⎪⎪ + ⎟ r ∂θ ⎥ ⎪⎪ ⎠ ⎦ 4 ⎭⎭
γθ ( p j +1 − p ) r Δθ h κ
⎡ ⎤ 3 ⎡ d ⎤⎥ vm ∂ ⎢ +γ vv α lva ρlva gθ + cwlv vl vl − ∑ ⎢cml ( vm − vl ) + cml ( vm − vl )⎥ ⎥ ⎢ ∂τ ⎦ m =1 ⎣ ⎢⎣ ⎥⎦ m ≠l ⎡ ⎤ 3 = γ vv ⎢ μ wlv ( vwl − vl ) − μ lwv ( vlw − vl ) + ∑ μ mlv ( vm − vl ) ⎥ + fν lv ⎢ ⎥ m =1 ⎢⎣ ⎥⎦ m ≠l
Introducing the geometry coefficients
βθ 1 =
rhκ,i −1 rhκ 1 1 , β = , βθ 3 = κ , βθ 4 = β θ 3 , βθ 5 = , βθ 6 = βθ 5 , θ2 r κ Δr r κ Δr r Δθ h Δz
12.17 Pipe networks: some basic definitions
575
the bv coefficients are defined as follows ⎧ 1 ⎫ bvl1 = − βθ 1 ⎨− min ⎡⎣0, (α l ρ l ul γ r )1 ⎤⎦ + (α l ρ lν lγ r )1 ⎬, Δ rh ⎭ ⎩ ⎧⎪ 1 ⎫⎪ bvl 2 = − βθ 2 ⎨ max ⎡⎣ 0, (α l ρl ul γ r )2 ⎤⎦ + (α l ρlν l γ r )2 ⎬, Δrh ,i −1 ⎭⎪ ⎩⎪ 1 ⎪⎫ ⎪⎧ bvl 3 = − βθ 3 ⎨− min ⎡⎣0, (α l ρl vl γ θ )3 ⎤⎦ + (α l ρlν l γ θ )3 κ ⎬, r Δθ j +1 ⎪⎭ ⎪⎩ ⎧ 1 ⎫ bvl 4 = − βθ 4 ⎨ max ⎡⎣ 0, (α l ρl vl γ θ )4 ⎤⎦ + (α l ρlν l γ θ )4 κ ⎬, r Δθ ⎭ ⎩ ⎧ 1 ⎫ bvl 5 = − βθ 5 ⎨− min ⎡⎣ 0, (α l ρl wl γ z )5 ⎤⎦ + (α l ρlν l γ z )5 ⎬, Δ zh ⎭ ⎩ ⎧⎪ 1 ⎫⎪ bvl 6 = − βθ 6 ⎨max ⎡⎣ 0, (α l ρl wl γ z )6 ⎤⎦ + (α l ρlν l γ z )6 ⎬. Δzh , k −1 ⎭⎪ ⎩⎪
Now we introduce the volume weighting coefficient C* =
V j +1 V + V j +1
.
This coefficient allows us easily to compute the mass flows entering the staggered cell through each face divided by the staggered cell volume by using the Blm − flux densities already computed for the non-staggered cells ⎛ Dν bvl , m = −C * ⎜ Blm − + β m l , m ⎜ ΔLh , m ⎝
⎞ * ⎟⎟ − 1 − C ⎠
(
⎛
) ⎜⎜ B ⎝
lm −
+ βm
Dlν, m ⎞ ⎟ . ΔLh , m ⎟⎠ j +1
With this abbreviation we have 6
γ vvα lva ρlva ( vl − vla ) / Δτ + ∑ bvl , m ( vl , m − vl ) ... m =1
Crrgpfkz"3409"Fkuetgvk|cvkqp"qh"vjg"czkcn"oqogpvwo"gswcvkqp" The z momentum equation discretized by using the donor cell concept is
γ vwα lwa ρlwa ( wl − wla ) / Δτ
576
12 Numerical solution methods for multi-phase flow problems
−
rhκ ⎧ 1 ⎫ ⎨− min ⎡⎣0, (α l ρl ul γ r )1 ⎤⎦ + (α l ρlν l γ r )1 ⎬ ( wli +1 − wl ) κ r Δr ⎩ Δrh ⎭
−
rhκ,i −1 ⎧⎪ 1 ⎫⎪ ⎨max ⎡⎣ 0, (α l ρl ul γ r )2 ⎤⎦ + (α l ρlν lγ r )2 ⎬ ( wli −1 − wl ) κ r Δr ⎩⎪ Δrh ,i −1 ⎭⎪
−
1 ⎧ 1 ⎫ ⎨ − min ⎡⎣ 0, (α l ρl vl γ θ )3 ⎤⎦ + (α l ρ lν lγ θ )3 κ ⎬ ( wlj +1 − wl ) r Δθ ⎩ r Δθ h ⎭
−
1 r Δθ
−
1 Δzh
−
γz 1 ⎧ 1 ⎫ ( pk +1 − p ) ⎨max ⎡⎣ 0, (α l ρl wl γ z )6 ⎤⎦ + (α l ρlν l γ z )6 ⎬ ( wlk −1 − wl ) +α lwa Δzh ⎩ Δz ⎭ Δz
κ
κ
⎧⎪ ⎫⎪ 1 ⎨ max ⎡⎣0, (α l ρl vl γ θ ) 4 ⎤⎦ + (α l ρlν l γ θ )4 κ ⎬ ( wlj −1 − wl ) r Δθ h , j −1 ⎭⎪ ⎩⎪
⎧ 1 ⎫ ⎨− min ⎡⎣0, (α l ρl wl γ z )5 ⎤⎦ + (α l ρlν l γ z )5 ⎬ ( wlk +1 − wl ) Δzk +1 ⎭ ⎩
⎡ ⎤ 3 ∂ ⎡ ⎤ +γ vw ⎢α lwa ρ lwa g z + cwlw wl wl − ∑ ⎢cmld ( wm − wl ) + cmlvm wm − wl ) ⎥ ⎥ ( ⎢ ∂τ ⎦⎥ m =1 ⎣ m≠l ⎣⎢ ⎦⎥ ⎡ ⎤ 3 ⎢ = γ vw μ wlw ( wwl − wl ) − μ lww ( wlw − wl ) + ∑ μ mlw ( wm − wl )⎥ + fν lw ⎢ ⎥ m =1 m ≠l ⎣⎢ ⎦⎥
Introducing the geometry coefficients
β z1 =
rhκ,i −1 rhκ 1 1 , β = , β z3 = κ , β z 4 = β z3 , β z5 = , β z6 = β z5 z2 r κ Δr r κ Δr r Δθ Δzh
the bw coefficients are defined as follows ⎧ 1 ⎫ bwl1 = − β z1 ⎨ − min ⎡⎣ 0, (α l ρl ul γ r )1 ⎤⎦ + (α l ρlν l γ r )1 ⎬, Δ rh ⎭ ⎩ ⎧⎪ 1 ⎫⎪ bwl 2 = − β z 2 ⎨max ⎡⎣ 0, (α l ρl ul γ r )2 ⎤⎦ + (α l ρlν l γ r )2 ⎬, Δrh ,i −1 ⎭⎪ ⎩⎪ ⎧ 1 ⎫ bwl 3 = − β z 3 ⎨− min ⎡⎣0, (α l ρl vl γ θ )3 ⎤⎦ + (α l ρ lν lγ θ )3 κ ⎬, r Δ θh ⎭ ⎩
12.17 Pipe networks: some basic definitions
577
⎧⎪ ⎫⎪ 1 bwl 4 = − β z 4 ⎨max ⎡⎣0, (α l ρl vl γ θ )4 ⎤⎦ + (α l ρlν l γ θ )4 κ ⎬, r Δθ h , j −1 ⎭⎪ ⎩⎪ ⎧ 1 ⎫ bwl 5 = − β z 5 ⎨− min ⎡⎣ 0, (α l ρl wl γ z )5 ⎤⎦ + (α l ρlν l γ z )5 ⎬, Δ z k +1 ⎭ ⎩ ⎧ 1 ⎫ bwl 6 = − β z 6 ⎨ max ⎡⎣ 0, (α l ρl wl γ z )6 ⎤⎦ + (α l ρlν lγ z )6 ⎬. Δz ⎭ ⎩ Now we introduce the volume weighting coefficient C* =
Vk +1 . V + Vk +1
This coefficient allows us easily to compute the mass flows entering the staggered cell through each face divided by the staggered cell volume by using the Blm − flux densities already computed for the non-staggered cells ⎛ Dν bwl , m = −C * ⎜ Blm − + β m l , m ⎜ ΔLh , m ⎝
⎞ * ⎟⎟ − 1 − C ⎠
(
⎛
) ⎜⎜ B ⎝
lm − + β m
Dlν, m ⎞ ⎟ . ΔLh , m ⎟⎠ k +1
With this abbreviation we have 6
γ vwα lwa ρlwa ( wl − wla ) / Δτ + ∑ bwl , m ( wl , m − wl ) ... m =1
Crrgpfkz"340:"Cpcn{vkecn"fgtkxcvkxgu"hqt"vjg"tgukfwcn"gttqt"qh"gcej" gswcvkqp"ykvj"tgurgev"vq"vjg"fgrgpfgpv"xctkcdngu" The velocities are assumed to be linearly dependent on the pressures, with the result that derivatives for the velocities with respect to pressure can be obtained directly from the momentum equations Vlmn = dVlmn − RVellm ( pm − p ) . The result is
∂ Vlmn ∂ Vlmn ∂ Vlmn = RVellm , = − RVellm , = 0 for x ≠ m , ∂p ∂ pm ∂ px and therefore
∂ blm + ∂V n = β mξlm + lm = β mξ lm + RVellm , ∂p ∂p
578
12 Numerical solution methods for multi-phase flow problems
∂ blm + ∂V n = β mξlm + lm = − β mξlm + RVellm , ∂ pm ∂ pm ∂ blm + = 0 for x ≠ m , ∂ px
∂ blm − ∂V n ∂ blm ∂ blm − = − β mξlm − lm = − β mξlm − RVellm , = , ∂p ∂p ∂p ∂p ∂ blm − ∂V n = − β mξlm − lm = β mξlm − RVellm , ∂ pm ∂ pm
∂ blm ∂ blm − = , ∂ pm ∂ pm
∂ blm − = 0 for x ≠ m , ∂ px
∂ blm = 0 for x ≠ m . ∂ px
Particle number density derivatives: Taking the derivatives of the LHS of the discretized particle number density conservation equation 6 1 f l n ≡ ⎡⎣ nl γ v − ( nl γ v )a ⎤⎦ / Δτ + ∑ ( blm + nl − blm − nl m ) − ( γ v + γ va ) Dnl 2 m =1
=
∂ fl n (n − n ) = 0 ∂ nl l l 0
we obtain: 6 ∂ fl n γ v = + ∑ blm + > 0 , ∂ nl Δτ m =1 6 ⎛ ∂b ∂ fl n ∂b = ∑ ⎜ nl lm + − nlm lm − ∂ pm x =1 ⎝ ∂ px ∂ px
⎞ ⎟ = − β m (ξlm + nl + ξlm − nlm ) RVellm , ⎠
for m = 1, 6, 6 6 ⎛ ∂b ∂ fl n ∂b ⎞ 6 ∂ fn = ∑ ⎜ nl lm + − nlm lm − ⎟ = ∑ ⎡⎣ β m (ξlm + nl + ξlm − nlm ) RVellm ⎤⎦ = − ∑ l . ∂ p m =1 ⎝ ∂ p ∂ p ⎠ m =1 m =1 ∂ pm
Mass conservation derivatives: Taking the derivatives of the LHS of the discretized mass conservation equation 6 1 f lα ≡ ⎡⎣α l ρl γ v − (α l ρl γ v )a ⎤⎦ / Δτ + ∑ [blm +α l ρl − blm − (α l ρl )m ] − ( γ v + γ va ) μl 2 m =1
=
∂ f lα (α − α l 0 ) = 0 ∂α l l
12.17 Pipe networks: some basic definitions
579
we obtain 6 ∂ f lα ∂ fl n ⎛γ ⎞ = ρl ⎜ v + ∑ blm + ⎟ ≡ ρl ; 0, ∂α l ∂ nl ⎝ Δτ m =1 ⎠ 6 ∂ f lα ∂ρ ⎛ γ ∂ρ ∂ fl n ⎞ α ∂ρ l ∂ flα = α l l ⎜ v + ∑ blm + ⎟ ≡ l ≡ αl l , ∂ sl ∂ sl ⎝ Δτ m =1 ∂ sl ∂ nl ⎠ ρl ∂ sl ∂α l 6 ∂ f lα ∂ρl ⎛ γ v ∂ρl ∂ f l n ⎞ α ∂ρl ∂ flα = αl + ∑ blm + ⎟ ≡ l ≡ αl , ⎜ ∂ Cil ∂ Cil ⎝ Δτ m =1 ∂ Cil ∂ nl ⎠ ρl ∂ Cil ∂α l 6 ∂ f lα ∂ρ ⎛ γ ∂b ∂b ⎤ ⎞ 6 ⎡ = α l l ⎜ v + ∑ blm + ⎟ + ∑ ⎢α l ρl lm + − (α l ρ l )m lm − ⎥ ∂p ∂ p ⎝ Δτ m =1 ∂p ∂p ⎦ ⎠ m =1 ⎣
=
α l ∂ f lα 6 + ∑ β m ⎡⎣ξlm +α l ρl + ξ lm − (α l ρ l ) m ⎤⎦ RVellm , ρl al2 ∂α l m =1
{
}
6 6 ⎡ ∂ f lα ∂b ∂b ⎤ ∂ρ = α l ρl ∑ lx + − ∑ ⎢(α l ρ l ) x lx − ⎥ − α lm blm − lm ∂ pm ∂ pm ⎦ ∂ pm x =1 ∂ pm x =1 ⎣
= − β m ⎡⎣ξlm +α l ρl + ξ lm − (α l ρl )m ⎤⎦ RVellm + β mα lmξlm −Vlmn
∂ρlm ∂ pm
⎡ ⎛ V n ∂ρlm ⎞ ⎤ = − β m ⎢ξlm +α l ρl RVellm + ξlm − (α l ρl )m ⎜ RVellm − lm ⎟ ⎥ for m = 1, 6. ρlm ∂ pm ⎠ ⎥⎦ ⎝ ⎣⎢
Specific entropy derivatives: Taking the derivatives of the LHS of the discretized entropy conservation equation f l s ≡ α a ρla γ va =
6 sl − sla 1 − ∑ [blm −α lm ρlm ( slm − sl ) ] + ( γ v + γ va ) μl+ sl − Dsl* Δτ 2 m =1
(
∂ fl s (s − s ) = 0 ∂ sl l l 0
we obtain: 6 ∂ fl s α la ρla γ va 1 = + ( γ v + γ v a ) μl+ + ∑ blm − (α l ρl )m ≥ 0 , ∂ sl Δτ 2 m =1 6 6 ∂ fl s ∂b = − ∑ (α l ρl )m ( slm − sl ) lm − = ∑ (α l ρl )m ( slm − sl ) β mξ lm − RVellm , ∂p ∂p m =1 m =1
)
580
12 Numerical solution methods for multi-phase flow problems 6 ∂ fl s ∂b ∂ρ = − ∑ (α l ρl ) x ( slx − sl ) lx − − blm −α lm ( slm − sl ) lm ∂ pm ∂ pm ∂ pm x =1
⎛ ∂ρ ⎞ = − β mξlm −α lm ( slm − sl ) ⎜ ρlm RVellm − Vlmn lm ⎟ for m = 1, 6. ∂ pm ⎠ ⎝
Inert mass conservation equation derivatives: Taking the derivatives of the LHS of the discretized inert mass conservation equation f ilC ≡ α la ρlaγ va +
6 ⎧⎡ Cil − Cila D* ⎪ −∑ ⎨ ⎢blm − (α l ρl ) m + β m ilm A ( Pem Δτ ΔLh , m m =1 ⎪ ⎢ ⎩⎣
⎤
)⎥ ( C ⎥⎦
il , m
⎪⎫ − Cil ) ⎬ ⎭⎪
∂ fC 1 (γ v + γ v a ) μl+ Cil − DCil = il (Cil − Cil 0 ) = 0 2 ∂ Cil
(
)
we obtain 6 ⎡ ⎤ ∂ f ilC γ vaα la ρla 1 D* = + ( γ v + γ va ) μl+ + ∑ ⎢blm − (α l ρl )m + β m ilm A ( Pem ) ⎥ ≥ 0 , ∂ Cil Δτ 2 ΔLh , m m =1 ⎣ ⎢ ⎦⎥ 6 6 ∂ f ilC ∂b = −∑ (α l ρl ) m ( Cilm − Cil ) lm = ∑ (α l ρl ) m ( Cilm − Cil ) β mξ lm − RVellm , ∂p ∂ p m =1 m =1 6 ∂ filC ∂b ∂ρ = −∑ (α l ρl ) x ( Cilx − Cil ) lx − − blm −α lm ( Cilm − Cil ) lm ∂ pm ∂ p ∂ pm x =1 m
⎛ ∂b ∂ρ ⎞ = −α lm ( Cilm − Cil ) ⎜ ρlm lm − + blm − lm ⎟ ∂ pm ∂ pm ⎠ ⎝ ⎛ ∂ρlm ⎞ = − β mξlm −α lm ( Cilm − Cil ) ⎜ ρlm RVellm − Vlmn ⎟ for m=1,6. ∂ pm ⎠ ⎝
Crrgpfkz"340;"Ukorng"kpvtqfwevkqp"vq"kvgtcvkxg"ogvjqfu"hqt"uqnwvkqp" qh"cnigdtcke"u{uvgou" Most of the iterative methods for solving the linear system of equations Ax = b
are based on the regular splitting of the coefficient matrix A = B − R.
References
581
All three matrices above are of the same range. The iteration method is then Bx n +1 = Rx n + b
or after eliminating R Bx n +1 = Bx n + b − Ax n .
Introducing the residual error vector at the previous solution r n = b − Ax n ,
we have a method for computing the next approximation x n +1 = x n + B −1r n ,
by adding to the previous estimate for the solution a correction vector Δx n , x n +1 = x n + Δ x n .
Δx n is in fact the solution of the algebraic system B Δx n = r n .
The matrix B has to be selected so as to enable non-expensive solution of the above system in terms of computer time and storage. Selecting
B = diag(A) results in the Jacobi method. Selection of B to be the lower off diagonal part including the diagonal leads to the Gauss-Seidel method. Selecting B to consists of non-expensively invertable blocks of A leads to the so called Block-Gauss-Seidel method. The secret of creating a powerful method is in appropriate selection of B. The reader will find valuable information on this subject in Saad (1996).
Tghgtgpegu" Addessio FL et al. TRAC-PF1 (February 1984) An advanced best-estimate computer program for pressurised water reactor analysis, NUREG/CR-3567, LA-9844-MS Addessio FL et al. (July-August 1985) TRAC-PF1/MOD1 Computer Code and Developmental Assessment, Nuclear Safety, vol 26 no 4 pp 438-454 Andersen JGM, Schang JC (1984) A predictor- corrector method for the BWR version of the TRAC computer code, AICHE Symposium Series, Heat Transfer - Niagara Falls, Farukhi NM (ed) 236.80, pp 275-280 Amsden AA (1985) KIVA: A computer program for two- and three-dimensional fluid flows with chemical reactions and fuel sprays, LA-10245-MS Bohl WR et al. (1988) Multi-phase flow in the advanced fluid dynamics model, ANS Proc. 1988, Nat. Heat Transfer Conf., July 24-July 27, Houston, Texas. HTC-3 pp 61-70 Caretto LS, Gosman AD, Patankar SV, Spalding DB (1973) Two calculation procedures for steady, three dimensional flow with recirculation, Proc. 3rd Int. Conf. on Numerical Methods in Fluid Mechanics, Springer, Lecture Notes in Physics, vol 11 no 19 pp 60-68
582
12 Numerical solution methods for multi-phase flow problems
Chow LC, Tien CL (1978) An examination of four differencing schemes for some elliptictype convection equations, Numerical Heat Transfer, vol 1 pp 87-100 Connell SD, Stow P (1986) A discussion and comparison of numerical techniques used to solve the Navier-Stokes and Euler equations, Int. J. for Num. Math. in Eng., vol 6 pp 155-163 Courant R, Isacson E, Rees M (1952) On the solution of non-linear hyperbolic differential equations by finite differentials, Commun. Pure Appl. Math., vol 5 p 243 Dearing JF (1985) A four-fluid model of PWR degraded cores, Third Int. Top. Meeting on Reactor Thermal Hydraulics, Newport, Rhode Island, Oct.15-18 1985, LA-UR- 85947/ CONF-85/007--3 Ferzinger JH and Paric (2002) Computational methods for fluid dynamics, Springer, 3rd Edition, Berlin Günther C (1988) Monotone Upwind-Verfahren 2.Ordnung zur Loesung der KonvektionsDiffusionsgleichungen, ZAMM. Z. angew. Math. Mech. vol 68 no 5, T383-T384 Günther C (August 1988) Vergleich verschiedener Differenzenverfahren zur numerischen Loesungen der 2-d Konvektions-Diffusionsgleichungen anhand eines Beispieles mit bekannter exakter Loesungen, Kernforschungszentrum Karlsruhe, KfK 4439 Gushchin VA, Shchennikov VV (1974) A monotonic difference scheme of second order accuracy, Zh. Vychisl. Mat. Mat. Fiz., vol 14 no 3 pp 789-792 Haaland SE (1984) Calculation of entrainment rate, initial values, and transverse velocity for the Patankar-Spalding method, Numerical Heat Transfer, vol 7 pp 39-57 Harlow FH, Amsden A A (1971) A numerical fluid dynamics calculations method for all flow speeds, J. Comp. Physics, vol 8 pp 197-213 Issa RI (1983) Numerical Methods for Two- and Tree-Dimensional Recirculating Flows, in Comp. Methods for Turbulent Transonic and Viscous Flow, Ed. Essers, J.A.: Hemisphere, Springer, pp 183 Kelly JM, Kohrt JR (1983) COBRA-TF: Flow blockage heat transfer program, Proc. Eleventh Water Reactor Safety Research Information Meeting, Gaithersburg Maryland, NUREG/CP-0048, 1 (Oct.24-28 1983) pp 209-232 Knight TD Ed. (1984) TRAC-PD2 Independent Assessment, NUREG/CR-3866, LA- 10166 Kolev NI (October 1986) IVA-2 A Computer Code for Modeling of Three Dimensional, ThreePhase, Three-Component Flow by Means of Three Velocity Fields in Cylindrical Geometry with Arbitrary Internal Structure Including Nuclear Reactor Core, Proc. of the Int. Seminar “Thermal physics 86” held in Rostock, German Democratic Republic, (in Russian) Kolev NI (Aug. 1987) A three field-diffusion model of three-phase, threecomponent flow for the transient 3D-computer code IVA2/001, Nuclear Technology, vol 78 pp 95-131 Kolev NI (1990) Derivatives for the state equations of multi-component mixtures for universal multi-component flow models, Nuclear Science and Engineering, vol 108 pp 74-87 Kolev NI (Sept. 1991) A three-field model of transient 3D multi-phase, three-component flow for the computer code IVA3, Part 2: Models for the interfacial transport phenomena. Code validation. KfK 4949 Kernforschungszentrum Karlsruhe Kolev NI , Tomiyama A, Sakaguchi T (Sept. 1991) Modeling of the mechanical interaction between the velocity fields in three phaseflow, Experimental Thermal and Fluid Science, vol 4 no 5 pp 525-545 Kolev NI (1993) Fragmentation and coalescence dynamics in multi-phase flows, Experimental Thermal and Fluid Science, vol 6 pp 211-251 Kolev NI (Apr. 23-27 1994a) IVA4 computer code: dynamic fragmentation model for liquid and its aplication to melt water interaction, Proc. ICONE-3, Third International Conf. on Nucl. Engineering, “Nuclear Power and Energy Future”, Kyoto, Japan. Presented at the “Workshop zur Kühlmittel/ Schmelze - Wechselwirkung”, 14-15 Nov. 1994, Cologne, Germany
References
583
Kolev NI (1994b) The code IVA4: Modeling of mass conservation in multi-phase multicomponent flows in heterogeneous porous media, Kerntechnik, vol 59 no 4-5 pp 226-237 Kolev NI (1994c) The code IVA4: Modeling of momentum conservation in multi-phase flows in heterogeneous porous media, Kerntechnik, vol 59 no 6 pp 249-258 Kolev NI (1995a) The code IVA4: Second Law of Thermodynamics for Multi-Phase MultiComponent Flows in Heterogeneous Porous Media, Kerntechnik vol 60 no 1 pp 28-39 Kolev NI (1995b) How accurate can we predict nucleate boiling, Experimental Thermal and Fluid Science, vol 10 pp 370-378 Kolev NI (1995c) The code IVA4: Nucleation and flashing model, Kerntechnik vol 60 (1995) no 6, pp 157-164. Also in: Proc. Second Int. Conf. On Multiphase Flow, Apr. 3-7, 1995, Kyoto; 1995 ASME & JSME Fluid Engineering Conference International Symposium on Validation of System Transient Analysis Codes, Aug. 13-18, 1995, Hilton Head (SC) USA; Int. Symposium on Two-Phase Flow Modeling and Experimentation, ERGIFE Place Hotel, Rome, Italy, October 9-11, 1995 Kolev NI (1995d) IVA4 computer code: The model for film boiling on a sphere in subcooled, saturated and superheated water, Proc. Second Int. Conference On Multiphase Flow, Apr. 3-7 1995, Kyoto, Japan. Presented At The “Workshop zur Kühlmittel/Schmelze - Wechselwirkung”, 14-15 Nov. 1994, Cologne, Germany Kolev NI (1996) Three fluid modeling with dynamic fragmentation and coalescence fiction or daily practice? 7th FARO Experts Group Meeting Ispra, October 15-16, 1996; Proceedings of OECD/CSNI Workshop on Transient thermal-hydraulic and neutronic codes requirements, Annapolis, Md, U.S.A., 5th-8th November 1996; 4th World Conference on Experimental Heat Transfer, Fluid Mechanics and Thermodynamics, ExHFT 4, Brussels, June 2-6, 1997; ASME Fluids Engineering Conference & Exhibition, The Hyatt Regency Vancouver, Vancouver, British Columbia, CANADA June 22-26, 1997, Invited Paper; Proceedings of 1997 International Seminar on Vapor Explosions and Explosive Eruptions (AMIGO-IMI), May 22-24, Aoba Kinen Kaikan of Tohoku University, Sendai-City, Japan Kolev NI (October 2-6, 2005) Strictly conservative limiter for fourth order CIP methods for multi-fluid analyses, The 11th International Topical Meeting on Nuclear Reactor Thermal-Hydraulics (NURETH-11) Log Number: 006, Popes Palace Conference Center, Avignon, France Köller A (1980) Anwendung numerischer Lösungsverfahren der Navier-Stokes-Gleichung zur Vermeidung von Wirbeln, Siemens Forsch.-u. Entwickl.-Ber., vol 9 no 2 pp 99-104 Latimaer BR, Pollard A (1985) Comparison of pressure-velocity coupling solution algorithms, Num. Heat Transfer, vol 8 pp 635-652 Leonard BP (1978) Third-order finite-difference method for steady two- dimensional convection, Num. Meth. in Laminar and Turbulent Flow, pp 807-819 Leonard BP, Mokhtari S (1990) Beyond first - order up winding: The ultra - sharp alternative for non - oscillatory Steady - state simulation of convection. Int. J. for Numerical Methods in Engineering, vol 30 pp 729-766 Leonard BP (1990) New flash: Upstream parabolic interpolation, Proc. 2nd GAMM Conference on Num. Meth. in Fluid Mechanics, Köln, Germany Liles D, Reed WR (1978) A semi-implicit method for two-phase fluid dynamics, J. of Comp. Physics, vol 26 pp 390-407 Liles D et al. (April 1981) TRAC-FD2 An advanced best-estimate computer program for pressurised water reactor loss-of-coolant accident analysis, NUREG/CR-2054, LA8709 MS Liles D, Mahaffy JM (1984) An approach to fluid mechanics calculations on serial and parallel computer architectures in large scale scientific computation, Parter SV (ed) Academic Press, Inc., Orlando, pp 141-159
584
12 Numerical solution methods for multi-phase flow problems
Mahaffy JH, Liles D (April 1979) Applications of implicit numerical methods to problems in two phase flow, NUREG/CR-0763, LA-7770-MS Mahaffy JH (1979) A stability-enhancing two-phase method for fluid flow calculations, NUREG/CR-0971, LA-7951-MS or J. of Comp. Physics, vol 46 no 3 (June 1983) Neuberger AW (March 1984) Optimierung eines numerischen Verfahrens zur Berechnung elliptischer Strömungen, DFVLR- FB84-16 Patankar SV, Spalding DB (1967) A finite-difference procedure for solving the equations of the two dimensional boundary layer, Int. J. Heat Mass Transfer vol 10 pp 1389-1411 Patankar SV (1978) A numerical method for conduction in composite materials, Flow in Irregular Geometries and Conjugate Heat Transfer, Proc. 6th Int. Heat Transfer Conf., Toronto, vol 3 p 297 Patankar SV (1980) Numerical heat transfer and fluid flow, Hemisphere, New York Patankar SV, Spalding DB (1972) A calculation procedure for heat, mass and momentum transfer in three dimensional parabolic flows, Int. J. Heat Mass Transfer, vol 15 pp 1787-1806 Patankar SV, Rafiinejad D, Spalding DB (1975) Calculation of the three-dimensional boundary layer with solution of all three momentum equations, Computer Methods in Applied Mechanics and Engineering, North-Holland Publ. Company, vol 6 pp 283-292 Patankar SV (1975) Numerical prediction of three dimensional flows, in Ed. Lauder B E, Studies in convection, theory, measurement and application, Academic Press, London, vol 1 Patankar SV, Basn DK, Alpay S (December 1977) A prediction of the three-dimensional velocity field of a deflected turbulent jet, Transactions of the ASME, Journal of Fluids Engineering, pp 758-767 Patankar S V (1981) A calculation procedure for two-dimensional elliptic situations, Numerical Heat Transfer, vol 4 pp 409-425 Prakash C (1984) Application of the locally analytic differencing scheme to some test problems for the convection- diffusion equation, Numerical Heat Transfer vol 7 pp 165-182 Patel MK, Markatos NC, Cross MS (1985) A critical evaluation of seven discretization schemes for convection-diffusion equations, Int. J. Num. Methods in Fluids, pp 225-244 Patel MK, Markatos NC (1986) An evaluation of eight discretization schemes for two- dimensional convection- diffusion equations, Int. J. for Numerical Methods in Fluids, vol 6 pp 129-154 Patel MK, Markatos NC (1986) An evaluation of eight discretization schemes for twodimensional convection-Diffusion equations, Int. J. for Numerical Methods in Fluids, vol 6 pp 129-154 Prior RJ (June 1979) Computational methods in thermal reactor safety, NUREG/CR-0851, LA-7856-MS Pollard A, Aiu LWA (1982) The calculation of some laminar flows using various discretization schemes, Comp. Math. Appl. Mesh. Eng., vol 35 pp 293-313 Rohatgi US (April 1985) Assessment of TRAC codes with dartmouth college countercurrent flow tests, Nuclear Technology vol 69 pp 100-106 Roscoe DF (1976) The solution of the three-dimensional Navier-Stokes equations using a new finite difference approach, Int. J. for Num. Math. in Eng., vol 10 pp 1299-1308 Saad Y (1996) Iterative methods for sparse linear systems, PWS Publishing Company, Boston, Sargis DA, Chan PC (1984) An implicit fractional step method for two-phase flow. Basic Aspects of Two Phase Flow and Heat Transfer, HTD-34 pp 127-136 Scarborough JB (1958) Numerical mathematical analysis, 4th ed., John Hopkins Press, Baltimore Spalding DB (1976) The Calculation of Free-Convection Phenomena in Gas-Liquid Mixtures, ICHMT Seminar Dubrovnik (1976), in Turbulent Buoyant Convection.
References
585
Eds. Afgan N, Spalding DB, Hemisphere, Washington pp 569- 586, or HTS Repport 76/11 (1977) Spalding DB (March 1979) Multi-phase flow prediction in power system equipment and components, EPRI Workshop on Basic Two-Phase Flow Modelling in Reactor Safety and Performance, Tampa, Florida Spalding DB (January 1980) Mathematical modeling of fluid mechanics, Heat-Transfer and Chemical - Reaction Processes, A lecture course, Imperial College of Science and Technology, Mech. Eng. Dep., London. HTS Report 80/1 Spalding DB (1980) Numerical computation of multi-phase fluid flow and heat transfer, in Recent Advances in Numerical Methods in Fluids, Eds. Taylor C, Morgan K, pp 139-167 Spalding DB (June 1981) Development in the IPSA procedure for numerical computation of multi-phase, slip, unequal temperature, etc, Imperial College of Science and Technology, Mech. Eng. Dep., London. HTS Report 81/11 Spalding DB (1981) A general purpose computer program for multi-dimensional one- and two-phase flow, Mathematics and Computers in Simulation vol XXIII pp 267-276 Spalding DB (1981) Numerical computation of multiphase flows, A course of 12 lectures with GENMIX 2P, Listing and 5 Appendices, HTS Report 81/8 Tanaka R, Nakamura T, Yabe T (2000) Constructing an exactly conservative scheme in a non conservative form, Comput. Phys. Commun., vol 126 pp 232 Tomiyama A (1990) Study on finite difference methods for fluid flow using difference operators and iteration matrices, PhD Thesis, Tokyo Institute of Technology Thurgood MJ, Cuta JM, Koontz AS, Kelly JM (1983) COBRA/TRAC - A thermal hydraulic code for transient analysis of nuclear reactor vessels and primary coolant systems, NUREG/CR-3046 1-5 van Doormaal JP, Raithby GD (1984) Enhancement of the SIMPLE method for prediction of incompressible fluid flows, Numerical Heat Transfer, vol 7 pp 147-163 Vanka SP (1985) Block-implicit calculation of steady turbulent recirculating flows, Int. J. Heat Mass Transfer, vol 28 no 11 pp 2093-2103 Williams KA, Liles DR (1984) Development and assessment of a numerical fluid dynamics model for non-equilibrium steam-water flows with entrained droplets, AICHE Symposium Series, Heat Transfer - Niagara Falls 1984, ed. by Farukhi NM, 236. 80 pp 416-425 Xiao F, Ikebata A and Hasegawa T (2004) Multi-fluid simulation by multi integrated moment method, Computers & Structures Yabe T, Takewaki H (Dezember 1986) CIP, A new numerical solver for general non linear hyperbolic equations in multi-dimension, KfK 4154, Kernforschungszentrum Karlsruhe Yabe T, Xiao F (1993) Description of complex sharp interface during shock wave interaction with liquid drop, Journal of Computational Physics of Japan, vol 62, no 8 pp 2537-2540 Yabe T, Xiao F, Mochizuki H (1995) Simulation technique for dynamic evaporation processes, Nuclear Engineering and Design, vol 155 pp 45-53 Yabe T, Xiao F, Utsumi T (2001) The constrained interpolation profile method for multiphase analysis, Journal of Computational Physics, vol 169 pp 556-593
35"Pwogtkecn"ogvjqfu"hqt"ownvk/rjcug"hnqy" kp"ewtxknkpgct"eqqtfkpcvg"u{uvgou"
This chapter presents a numerical solution method for multi-phase flow analysis based on local volume and time averaged conservation equations. The emphasis of this development was to create a computer code architecture that absorb all the constitutive physics and functionality from the past 25years development of the three fluid multi-component IVA-entropy concept for multiphase flows into a boundary fitted orthogonal coordinate framework. Collocated discretization for the momentum equations is used followed by weighted averaging for the staggered grids resulting in analytical expressions for the normal velocities. Using the entropy concept analytical reduction to a pressure-velocity coupling is found. The performance of the method is demonstrated by comparison of two cases for which experimental results and numerical solution with the previous method are available. The agreement demonstrates the success of this development.
3503"Kpvtqfwevkqp" We extend now the method described in the previous chapter to more arbitrary geometry. Instead of considering Cartesian or cylindrical geometry only, we will consider an integration space called a block in which the computational finite volumes fit inside the block so that the outermost faces of the external layer of the finite volumes create the face of the block. Similarly, bodies immersed into this space have external faces identical with the faces of the environmental computational cells. Such blocks can be inter-connected. With this technology multiphase flows in arbitrary interfaces can be conveniently handled. For understanding the material presented in this section I strongly recommend going over Appendixes 1 and 2 before continuing reading. Before starting with the description of the new method let us summarize briefly the state of the art in this field. In the last ten years the numerical modeling of single-phase flow in boundary fitted coordinates is becoming standard in the industry. This is not the case with the numerical modeling of multiphase flows. There are some providers of singlephase-computer codes claming that their codes can simulate multi-phase flows. Taking close looks of the solution methods of these codes reveals that existing single phase solvers are used and a provision is given to the user to add an other
588
13 Numerical methods for multi-phase flow in curvilinear coordinate systems
velocity field and define explicit the interfacial interaction physics. This strategy does not account for the feed back of the strong interfacial interactions on the mathematical solution methods - see the discussion in Miettinen and Schmidt (2002). Multi-phase flow simulations require specific solution methods accounting for this specific physics - see for instance the discussion by Antal et al. (2000). There are groups of methods that are solving single phase conservation equations with surface tracking, see the state of the art part of Tryggavson et al. (2001) work. This is in fact a direct numerical simulation that is outside of the scope of this chapter. To mention few of them: In Japan a powerful family of cubic-interpolation methods (CIP) is developed based on the pseudo-characteristic method of lines Takewaki, Nishiguchi and Yabe (1985), Takewaki and Yabe (1987), Nakamura, Tanaka, Yabe and Takizawa (2001), Yabe, Xiao and Utsumi (2001), Yabe, Tanaka, Nakamura and Xiao (2001), Yabe, Xiao and Utsumi (2001), Yabe and Takei (1988), Xiao and Yabe (2001), Xiao, Yabe, Peng and Kobayashi (2002), Xiao (2003), Xiao and Ikebata (2003), Yabe and Wang (1991), Yabe and Aoki (1991), Yabe et al. (1991). In USA particle tracking and level-set surface tracking methods are very popular; see for instance Sussman, Smereka and Oslier (1994), Osher and Fredkiw (2003), Swthian (1996), Tryggavson at al. (2001). The third group of DNS method with surface tracking is the latticeBoltzman family, see Hou et al. (1995), Nourgaliev et al. (2002) and the references given there. To the family of emerging methods the so called diffuse interface methods based on high order thermodynamics can be mentioned, see Verschueren (1999), Jamet et al. (2001). Let as emphasize once again, that unlike those methods, our work is concentrated on methods solving the local volume and time averaged multiphase flow equations which are much different then the single phase equations. In Europe two developments for solving two-fluid conservation equations in unstructured grids are known to me. Staedke et al. (1998) developed a solution method based on the method of characteristics using unstructured grid in a single domain. The authors added artificial terms to enforce hyperbolicy in the initially incomplete system of partial differential equations that contain derivatives which do not have any physical meaning. Toumi et al. (2000) started again from the incomplete system for two-fluid two phase flows without interaction terms and included them later for a specific class of processes; see Kumbaro et al. (2002). These authors extended the approximate Riemann solver originally developed for single phase flows by Roe to two-fluid flows. One application example of the method is demonstrated in a single space domain in Toumi et al. (2000), Kumbaro et al. (2002). No industrial applications of these two methods have been reported so far. One should note that it is well know that if proper local volume averaging is applied the originating interfacial interaction terms provide naturally hyperbolicy of the system of PDS and there is no need for artificial terms without any physical meaning. An example for the resolution of this problem is given by van Wijngaarden in 1976 among many others. In USA Lahey and Drew demonstrated clearly in 1999 haw by careful elaboration of the constitutive relationships starting from first principles variety of steady state processes including frequency dependent acoustics can be successfully simulated. Actually, the idea by Lahey and Drew (1999) is a further development
13.2 Nodes, grids, meshes, topology - some basic definitions
589
of the proposal made by Harlow and Amsden in 1975 where liquid (1) in vapor (2) and vapor (3) in liquid (4) are grouped in two velocity fields 1 + 2 and 3 + 4. The treatment of Lahey and Drew (1999) is based on four velocity fields. Antal et al. (2000) started developing the NPHASE multi-domain multi-phase flows code based on the single phase Rhie and Chow numerical method extended to multiphase flows. Application example is given for T-pipe bubble flows with 10 groups of bubble diameters. Two works that can be considered as a subset of this approach are reported in Tomiyama et al. (2000) Gregor, Petelin and Tiselj (2000). Another direction of development in USA that can be observed is the use of the volume of fluid method with computing the surface tension force as a volumetric force Hirt (1993), Kothe et al. (1996), Brackbill et al. (1992), Rider and Kothe (1998).
3504"Pqfgu."itkfu."ogujgu."vqrqnqi{"/"uqog"dcuke" fghkpkvkqpu"" Database concept: Let as consider the data base concept. The data volume is made up of points, or nodes, which themselves define in their neighborhood a volume element. We use hexahedrons (Fig. 4 in Appendix 1). A hexahedron is a 3D volume element with six sides and eight vertices. The vertices are connected in an order that mimics the way nodes are numbered in the data volume. The nodes in the data volume are numbered by beginning with 1 at the data volume’s origin. Node numbering increases, with x changing fastest. This means node numbering increases along the x axis first, the y axis second, and the z axis last, until all nodes are numbered. The numbering of the vertices of the volume elements follows the same rule. It starts with the vertex being closest to the data volume’s origin, moves along x, then y, and then z. Grid: A grid is a set of locations in a 3D data volume defined with x, y, and z coordinates. The locations are called nodes, which are connected in a specific order to create the topology of the grid. A grid can be regular or irregular depending on how its nodes are represented as points. A regular grid’s nodes are evenly spaced in x, y, and z directions, respectively. A regular grid’s nodes are specified with x, y, and z offsets from the data volume’s origin. A regular grid may have equidistant or non-equidistant spacing. If equidistant spacing is used all areas of the data volumes have the same resolution. This suits data with regular sample intervals. An irregular grid’s nodes need not be evenly spaced or in a rectangular configuration. This suits data with a specific area of interest that require finer sampling. Because nodes may not be evenly spaced, each node’s xyz coordinates must be explicitly listed. Topology: A topology defines an array of elements by specifying the connectivity of the element’s vertices or nodes. It builds a volume from separate elements by specifying how they are connected together. The elements can be
590
13 Numerical methods for multi-phase flow in curvilinear coordinate systems
3D volume elements or 2D surface elements. A topology is either regular or irregular, depending upon what types of elements it defines and how they are structured. A regular topology defines a data volume’s node connectivity. We assume a hexahedron volume element type. A regular topology can be used for regular or irregular grids. An irregular topology defines the node connectivity of either a data volume or geometric surface elements. An irregular topology data volume can be composed of either hexahedron or tetrahedron volume elements. Geometry objects are composed of points, lines, or polygons. An element list has to explicitly specify how the nodes connect to form these elements. Volume elements: The volume elements are the smallest building blocks of a data volume topology. Mesh: A mesh is a grid combined with specific topology for the volume of data. We distinguish the following mesh types: regular meshes, irregular or structured meshes, unstructured meshes, and geometry meshes. A regular mesh consists of grid having regular spacing and regular topology that consists of simple, rectangular array of volume elements. An irregular mesh explicitly specifies the xyz coordinates of each node in a node list. As in the regular mesh, the topology is regular, although individual elements are formed by explicit xyz node locations. The grid may be irregular or rectilinear. An unstructured mesh explicitly defines the topology. Each topology element is explicitly defined by its node connectivity in an element list. The grid may be regular or irregular. In this sense we are dealing in this Chapter with multi-blocks each of them consisting of • •
irregular grid’s nodes, irregular meshes, regular topology with hexahedron volume element type.
The integration space is built by a specified number of interconnected blocks.
3505"Hqtowncvkqp"qh"vjg"ocvjgocvkecn"rtqdngo" Consider the following mathematical problem: A multi-phase flow is described by the following vector of dependent variables UT = (α m , T1 , s2 , s3 , Cil , nl , p, ul , vl , wl ) ,
where l = 1, 2, 3 , i1 = 1...n1 , i 2 = 1...n 2 , i3 = 1...n3
which is a function of the three space coordinates ( x, y, z ) , and of the time τ ,
13.3 Formulation of the mathematical problem
591
U = U ( x, y , z , τ ) .
The relationship U = U( x, y , z ,τ ) is defined by the volume-averaged and successively time-averaged mass, momentum and energy conservation equations derived in Chapters 1, 2, 5 Kolev (1994a, b, 1995, 1997, 1998) as well as by initial conditions, boundary conditions, and geometry. The conservation equations are transformed in a curvilinear coordinate system ξ , η , ζ as shown in Chapter 11, Kolev (2001). The flux form of these equations is given in Chapter 11. As shown in Chapter 11, Kolev (2001), the conservation principles lead to a system of 19+n1+n2+n3 non-linear, non-homogeneous partial differential equations with variable coefficients. This system is defined in the three-dimensional domain R. The initial conditions of U(τ = 0) = U a in R and the boundary conditions acting at the interface separating the integration space from its environment are given. The solution required is for conditions after the time interval Δτ has elapsed. The previous time variables are assigned the index a. The time variables not denoted with a are either in the new time plane, or are the best available guesses for the new time plane. In order to enable modeling of flows with arbitrary obstacles and inclusions in the integration space as is usually expected for technical applications, surface permeabilities are defined (γ ξ , γ η , γ ζ ) = functions of (ξ ,η , ζ ,τ ) ,
at the virtual surfaces that separate each computational cell from its environment. By definition, the surface permeabilities have values between one and zero, 0 ≤ each of all (γ ξ , γ η , γ ζ ) ‘s ≤ 1 .
A volumetric porosity
γ v = γ v (ξ ,η , ζ , τ ) is assigned to each computational cell, with 0 < γv ≤1.
The surface permeabilities and the volume porosities are not expected to be smooth functions of the space coordinates in the region R and of time. For this reason, one constructs a frame of geometrical flow obstacles which are functions of space and time. This permits a large number of extremely interesting technical applications of this type of approach. In order to construct useful numerical solutions it is essential that an appropriate set of constitutive relations be available: state equations, thermodynamic derivatives, equations for estimation of the transport properties, correlations modeling the heat, mass and momentum transport across the surfaces dividing the separate velocity fields, etc. These relationships together are called closure equations. This very complex problem will not be discussed in Volume II. Only the numerics will be addressed here.
592
13 Numerical methods for multi-phase flow in curvilinear coordinate systems
3506"Fkuetgvk|cvkqp"qh"vjg"ocuu"eqpugtxcvkqp"gswcvkqpu" 350603"Kpvgitcvkqp"qxgt"c"hkpkvg"vkog"uvgr"cpf"hkpkvg"eqpvtqn"xqnwog" We start with the conservation equation (10.56) for the species i inside the velocity field l in the curvilinear coordinate system ∂ ∂ ∂ α l ρl Cil g γ v + γ ξ g a1 ⋅ G il + γ η g a 2 ⋅ G il ∂ξ ∂η ∂τ ∂ + γ ζ g a3 ⋅ G il = γ v g μil , ∂ζ
(
(
)
(
)
(
)
)
(13.1)
where the species mass flow rate vector is defined as follows ⎡ ⎛ ∂C ∂C ∂C ⎞ ⎤ G il = α l ρl ⎢Cil ( Vl − Vcs ) − Dil* ⎜ a1 il + a2 il + a3 il ⎟ ⎥ . ∂η ∂ζ ⎠ ⎦ ⎝ ∂ξ ⎣
(13.2)
Note that for Cil = 1 we have the mass flow vector of the velocity field G l = α l ρl ( Vl − Vcs ) .
(13.3)
Next we will use the following basic relationships from Appendix 2 between the surface vectors and the contravariant vectors, and between the Jacobian determinant and the infinitesimal spatial and volume increments g a1 =
S1 , ∂η∂ζ
g a2 =
S2 , ∂ξ∂ζ
g a3 =
S3 , ∂ξ∂η
g=
dV . dξ dη dζ
(13.4-7)
We will integrate both sides of the equation over the time, and spatial intervals ∂τ , ∂ξ , ∂η , and ∂ζ respectively. We start with the first term ⎡
∫∫∫ ⎢⎣ ∫ ΔV
Δτ
0
∂ ⎤ ⎡ Δτ ∂ ⎤ α l ρl Cil g γ v ∂τ ⎥∂ξ∂η∂ζ = ⎢ ∫ α l ρl Cilγ v ) ∂τ ⎥ ∫∫∫ dV ( 0 ∂τ ∂τ ⎦ ⎣ ⎦ ΔV
(
)
= ⎡(α l ρl Cil γ v ) − (α l ρl Cil γ v ) ⎤ ΔV . a⎦ ⎣
(13.8)
This result is obtained under the assumption that there is no spatial variation of the properties (α l ρl Cilγ v ) inside the cell. The integration of the other terms gives the following results: ∂
∫ ∫∫∫ ∂ξ (γ ξ Δτ
0
ΔV
(
)
g a1 ⋅ G il ∂ξ∂η∂ζ ∂τ = Δτ ∫
) (
= Δτ ⎡ γ ξ S1 ⋅ G il − γ ξ S1 ⋅ G il ⎣
Δξ
0
)
i −1
⎤ ⎦
∂ (γ ξ S1 ⋅ G il ) ∂ξ ∂ξ
13.4 Discretization of the mass conservation equations
S ⎡ S = ΔτΔV ⎢γ ξ 1 e1 ⋅ G il − γ ξ ,i −1 2 e1 ⋅ G il ΔV ⎣ ΔV
(
∂
∫ ∫∫∫ ∂η (γ η Δτ
0
ΔV
(
)
(
)
)
Δη
g a 2 ⋅ G il ∂ξ∂η∂ζ ∂τ = Δτ ∫
0
) (
= Δτ ⎡ γ η S 2 ⋅ G il − γ η S 2 ⋅ G il ⎣⎢
)
j −1
∂
∫ ∫∫∫ ∂ζ (γ ζ Δτ
0
ΔV
(
)
(
)
g a3 ⋅ G il ∂ξ∂η∂ζ ∂τ = Δτ ∫
) (
Δζ
)
k −1
Δτ
∫ ∫∫∫ γ 0
ΔV
v
(13.9)
∂ (γ η S2 ⋅ G il ) ∂η ∂η
⎤ ⎥, ⎦
j −1
(13.10)
∂ (γ ζ S3 ⋅ G il ) ∂ζ ∂ζ
⎤ ⎦
S S ⎡ = ΔτΔV ⎢γ ζ 5 e3 ⋅ G il − γ ζ ,k −1 6 e3 ⋅ G il ΔV ⎣ ΔV
(
)
0
= Δτ ⎡ γ ζ S 3 ⋅ G il − γ ζ S 3 ⋅ G il ⎣
⎤ ⎥, ⎦
⎤ ⎦⎥
S ⎡ S = ΔτΔV ⎢γ η 3 e 2 ⋅ G il − γ η , j −1 4 e 2 ⋅ G il ΔV ⎣ ΔV
(
i −1
593
)
(
)
k −1
⎤ ⎥, ⎦
(13.11)
g μil ∂ξ∂η∂ζ ∂τ = Δτ ∫∫∫ γ v μ il dV = Δτγ v μil ΔV .
(13.12)
ΔV
It is convenient to introduce the numbering at the surfaces of the control volumes 1 to 6 corresponding to high-i, low-i, high-j, low-j, high-k and low-k, respectively. We first define the unit surface vector ( e ) at each surface m as outwards directed: m
(e)
1
= e1 , ( e ) = −e1i−1 , ( e ) = e 2 , ( e ) = −e 2j −1 , ( e ) = e3 , ( e ) = −e3k −1 . (13.13-18) 2
3
4
5
6
With this we have a short notation of the corresponding discretized concentration conservation equation 6
α l ρl Cilγ v − α la ρla Cilaγ va + Δτ ∑ β m ( e ) ⋅ G il ,m = Δτγ v μ il . m
(13.19)
m=1
We immediately recognize that it is effective to compute once the geometry coefficients S S1 S S , β 2 = γ ξ ,i −1 2 , β 3 = γ η 3 , β 4 = γ η , j −1 4 , ΔV ΔV ΔV ΔV S5 S6 β5 = γ ζ , β 6 = γ ζ , k −1 , ΔV ΔV
β1 = γ ξ
(13.20-25)
before the process simulation, to store them, and to update only those that change during the computation. Secondly, we see that these coefficients contain exact physical geometry information. Note that for cylindrical coordinate systems we have
594
13 Numerical methods for multi-phase flow in curvilinear coordinate systems
β1 = β5 =
rhκ γ r κ
r Δr
γz Δz
, β2 =
, β6 =
(rhκ γ r )i −1 κ
r Δr
γ z , k −1 Δz
, β3 =
γθ
κ
r Δθ
, β4 =
γ θ , j −1 r κ Δθ
,
,
(13.26-31)
and for Cartesian setting κ = 0 and r = x , θ = y ,
β1 =
γx Δx
, β2 =
γ x ,i −1 Δx
, β3 =
γy Δy
, β4 =
γ y , j −1 Δy
, β5 =
γz Δz
, β6 =
γ z ,k −1 Δz
, (13.32-39)
compare with Section 11.4. Setting Cil = 1 in Eq. (13.19) we obtain the discretized mass conservation equation of each velocity field 6
α l ρlγ v − α la ρlaγ va + Δτ ∑ β m ( e ) ⋅ G l ,m = Δτγ v μ l . m
(13.40)
m =1
Next we derive the useful non-conservative form of the concentration equations. We multiply Eq. (13.40) by the concentration at the new time plane and subtract the resulting equation from Eq. (13.19). Then the field mass source term is split in two non-negative parts μ l = μ l+ + μ l− . The result is 6
α la ρla ( Cil − Cila ) γ va + Δτ ∑ β m ( e ) ⋅ ( G il ,m − Cil G l ,m ) + Δτγ v μl+ Cil = Δτγ v DCil . m
m =1
(13.41) where DCil = μ il + μ Cil . Note that − l
G il ,m − Cil G l ,m ⎡ ⎛ ∂C ∂C ∂C ⎞ ⎤ = ⎡⎣α l ρl ( Vl − Vcs ) ⎤⎦ m ( Cil ,m − Cil ) − ⎢α l ρl Dil* ⎜ a1 il + a2 il + a3 il ⎟ ⎥ . ∂η ∂ζ ⎠ ⎦ m ⎝ ∂ξ ⎣ (13.42) Up to this point of the derivation we did not made any assumption about the computation of the properties at the surfaces of the control volume.
350604"Vjg"fqpqt/egnn"eqpegrv" The concept of the so called donor-cell for the convective terms is now introduced. Flow of given scalars takes the values of the scalars at the cell where the flow is coming from. Mathematically it is expressed as follows. First we define velocity normal to the cell surfaces and outwards directed Vl n,m = ( e ) ⋅ ( Vl − Vcs )m , m
(13.43)
then the switch functions (to store them use signet integers in computer codes, it saves memory)
13.4 Discretization of the mass conservation equations
ξlm+ =
1⎡ 1 + sign Vlmn ⎤ , ⎦ 2⎣
( )
595
(13.44)
ξlm− = 1 − ξlm+ ,
(13.45)
and then the b coefficients as follows blm+ = β mξlm+Vlmn ≥ 0 ,
(13.46)
blm− = − β mξlm−Vlmn ≥ 0 .
(13.47)
If the normal outwards directed velocity is positive the +b coefficients are unity and the –b coefficients are zero and vice versa. In this case the normal mass flow rate at the surfaces is
(e)
m
⋅ G l ,m = ⎡⎣α l ρl e ⋅ ( Vl − Vcs ) ⎤⎦ m = (ξlm+α l ρl + ξlm−α l ,m ρl ,m ) Vlmn .
(13.48)
Using the above result the mass conservation equations for each field result is 6
α l ρlγ v − α la ρlaγ va + Δτ ∑ ( blm+α l ρ l − Blm− ) − Δτγ v μ l = 0 .
(13.49)
m =1
In the donor-cell concept the term Blm− = blm−α l ,m ρl ,m
(13.50)
plays an important role. Blm− is in fact the mass flow entering the cell from the face m divided by the volume of the cell. Once computed for the mass conservation equation it is stored and used subsequently in all other conservation equations. At this point the method used for computation of the field volumetric fractions by iteration using the point Gauss-Seidel method for known velocity vectors and thermal properties will be described. Consider the field variables α l ρl in the convective terms associated with the output flow in the new time plane, and α lm ρlm in the neighboring cells m as the best available guesses for the new time plane. Solving Eq. (13.49) with respect to α l ρl gives ⎡ ⎛ α ρ α l ρl = ⎢γ va ⎜ μl + la la Δτ ⎣ ⎝
⎤ ⎞ 6 ⎟ + ∑ Blm− ⎥ ⎠ m=1 ⎦
6 ⎛ γv ⎞ + ∑ blm+ ⎟ . ⎜ ⎝ Δτ m=1 ⎠
(13.51)
Here 6 γv + ∑ blm+ > 0 Δτ m=1
(13.52)
596
13 Numerical methods for multi-phase flow in curvilinear coordinate systems
is ensured because γ v is not allowed to be zero. For a field that is just originating we have 6
αl =
Δτ
ρl
γ va μl + ∑ Blm− m =1 6
γ v + Δτ ∑ blm+
.
(13.53)
m =1
6
Obviously the field can originate due to convection,
∑B m =1
lm −
> 0 , or due to an in-
cell mass source, μ l > 0 , or due to the simultaneous appearance of both phenomena. In case of origination caused by in-cell mass source terms it is important to define the initial density, ρl , in order to compute α l = Δτμl / ρl . The best mass conservation in such procedures is ensured if the following sequence is used for computation of the volume fractions:
α 2 = α 2 ρ 2 / ρ 2 , α 3 = α 3 ρ 3 / ρ 3 , α1 = 1 − α 2 − α 3 .
(13.54-56)
For designing the pressure-velocity coupling the form of the discretized mass conservation is required that explicitly contains the normal velocities, 6
(α l ρlγ v − α la ρlaγ va ) / Δτ + ∑ β m (ξ lm+α l ρl + ξ lm−α lm ρ lm ) Vlmn − γ v μ l = 0 . m =1
(13.57) The mass flow rate of the species i inside the field l at the cell surface m is then
(e)
m
⎡ ⎛ ∂C ∂C ∂C ⎞ ⎤ ⋅ G il ,m = ⎡⎣α l ρl Cil e ⋅ ( Vl − Vcs ) ⎤⎦ − ⎢α l ρl Dil* e ⋅ ⎜ a1 il + a 2 il + a3 il ⎟ ⎥ m ∂η ∂ζ ⎠ ⎦ m ⎝ ∂ξ ⎣
⎡ ⎛ ∂C ∂C ∂C ⎞ ⎤ = (ξlm +α l ρl Cil + ξlm−α l ,m ρl ,mCil ,m ) Vlmn − ⎢α l ρl Dil*e ⋅ ⎜ a1 il + a 2 il + a3 il ⎟ ⎥ , ∂η ∂ζ ⎠ ⎦ m ⎝ ∂ξ ⎣ (13.58)
and consequently
(e)
m
⋅ ( G il ,m − Cil G l ,m )
⎡ ⎛ ∂C ∂C ∂C ⎞ ⎤ = ξlm−α l ,m ρl ,mVlmn ( Cil ,m − Cil ) − ⎢α l ρl Dil* e ⋅ ⎜ a1 il + a 2 il + a3 il ⎟ ⎥ . (13.59) ∂η ∂ζ ⎠ ⎦ m ⎝ ∂ξ ⎣
Thus Eq. (13.41) takes the intermediate form
13.4 Discretization of the mass conservation equations
α la ρla ( Cil − Cila ) γ va
(
)
597
⎧ ⎫ ⎪ ⎪ ⎪ Blm− Cil ,m − Cil ⎪ 6 ⎪ ⎪ − Δτ ∑ ⎨ ⎬ m =1 ⎪ ⎪ ⎪+ β ⎡α ρ D* e ⋅ ⎛ a1 ∂Cil + a 2 ∂Cil + a3 ∂Cil ⎞ ⎤ ⎪ ⎢ ⎥ ⎜ ⎟ ⎪ m ⎢ l l il ⎝ ∂ξ ∂η ∂ζ ⎠ ⎥⎦ m ⎪ ⎣ ⎩ ⎭
(
)
= Δτγ v μil − Cil μl .
(13.60)
350605"Vyq"ogvjqfu"hqt"eqorwvkpi"vjg"hkpkvg"fkhhgtgpeg"crrtqzkocvkqpu" qh"vjg"eqpvtcxctkcpv"xgevqtu"cv"vjg"egnn"egpvgt" The contravariant vectors for each particular surface can be expressed by a1 =
S1 S2 S3 ∂ξ , a 2 = ∂η , a3 = ∂ζ . dV dV dV
(13.61-63)
Note that the contravariant vectors normal to each control volume surface are conveniently computed for equidistant discretization in the computational space as follows
(a )
=
( S )1 S1 1 S1 = = (e) , ΔV 1 ΔV 1 ΔV 1
(a )
(a )
=
( S )3 S 3 3 S2 = = (e) , ΔV 3 ΔV 3 ΔV3
(a )
=
( S ) 5 S5 5 S3 = = (e ) , ΔV 5 ΔV 5 ΔV5
(a )
1
1
2
(a )
3
3
5
1
2
2
4
3
6
=
( S )2 S1 S 2 =− = − 2 ( e ) , (13.64-65) ΔV 2 ΔV 2 ΔV 2
=
( S )4 S2 S 4 =− = − 4 ( e ) , (13.66-67) ΔV 4 ΔV 4 ΔV 4
=
( S )6 S S3 6 =− = − 6 ( e ) , (13.68-69) ΔV 6 ΔV 6 ΔV 6
where the volume associated with these vectors is ΔV m =
1 ( ΔV + ΔVm ) . 2
(13.70)
The finite volume method: There are two practicable methods for approximation of the contravariant vectors at the cell center. The first one makes use of the already computed normal interface vectors in the following way: a1c =
1⎡ 1 a 2⎣
( ) + (a ) ⎤⎦ , a 1
1
2
2 c
=
1⎡ 2 a 2⎣
( ) + ( a ) ⎤⎦ , a 2
3
4
3 c
=
1⎡ 3 a 2⎣
( ) + ( a ) ⎤⎦ 3
5
6
(13.71-73)
598
13 Numerical methods for multi-phase flow in curvilinear coordinate systems 7 8 5
ζ
z
3
4
6
η
y 1
2
ξ x
Fig. 13.1 Numbering of the vertices
The finite difference method: The second method uses the coordinates of the vertices of the control volume directly, Fig. 13.1. First we define the position at the cell surfaces that will be used to compute the transformation metrics as follows: rS1 =
1 ( r2 + r4 + r8 + r6 ) , 4
rS 2 =
1 ( r1 + r3 + r7 + r5 ) , 4
(13.74-75)
rS 3 =
1 ( r3 + r4 + r8 + r7 ) , 4
rS 4 =
1 ( r1 + r2 + r6 + r5 ) , 4
(13.76-77)
rS 5 =
1 ( r5 + r6 + r8 + r7 ) , 4
rS 6 =
1 ( r1 + r2 + r4 + r3 ) . 4
(13.78-79)
Then we compute the inverse metrics of the coordinate transformation for equidistant discretization in the transformed space ⎛ ∂x ⎜ ⎜ ∂ξ ⎜ ∂y ⎜ ⎜ ∂ξ ⎜ ∂z ⎜ ⎝ ∂ξ
∂x ∂η ∂y ∂η ∂z ∂η
∂x ∂ζ ∂y ∂ζ ∂z ∂ζ
⎞ ⎟ ⎟ ⎛ xs1 − xs 2 ⎟ ⎜ ⎟ = ⎜ ys1 − ys 2 ⎟ ⎜z −z ⎟ ⎝ s1 s 2 ⎟ ⎠
xs 3 − xs 4 ys 3 − ys 4 zs 3 − zs 4
xs 5 − xs 6 ⎞ ⎟ ys 5 − ys 6 ⎟ . zs 5 − zs 6 ⎟⎠
(13.80)
Then we compute the Jacobian determinant and the metrics of the coordinate transformation for equidistant discretization in the transformed space. As already mentioned all this information belongs to the center of the cell. However, the off-diagonal geometry information is required at the cell surfaces. For both cases we use the two corresponding neighbor vectors to compute the contravariant vectors at the cell surfaces as follows
13.4 Discretization of the mass conservation equations
599
1 2 ( ac + ac2,i+1 ) , ( a3 )1 = 12 ( a3c + a3c,i+1 ), 2 ( a2 )2 = 12 ( ac2 + ac2,i−1 ) , ( a3 )2 = 12 ( a3c + a3c,i−1 ) , ( a1 )3 = 12 ( a1c + a1c, j+1 ) , (a3 )3 = 12 ( a3c + a3c, j+1 ),
(a )
=
(a )
4
=
(a )
5
(a )
6
2
1
1
1
1
1 1 1 ac + a1c , j −1 ) , ( a 3 ) = ( a 3c + ac3, j −1 ), ( 4 2 2
1 1 ( ac + a1c,k +1 ) , (a2 )5 = 12 ( ac2 + ac2,k +1 ) , 2 1 1 = ( a1c + a1c ,k −1 ) , ( a 2 ) = ( a 2c + a c2,k −1 ) . 6 2 2
=
(13.81-94)
350606"Fkuetgvk|cvkqp"qh"vjg"fkhhwukqp"vgtou" 13.4.4.1 General Our next task is to find appropriate finite difference approximation for the six diffusion terms
⎡ ⎛ 1 ∂Cil ∂C ∂C ⎞ ⎤ * + a 2 il + a3 il ⎟ ⎥ , ⎢α l ρl Dil e ⋅ ⎜ a ∂η ∂ζ ⎠ ⎦ m ⎝ ∂ξ ⎣ The geometric properties computed by using the control volume approach in the previous section are used to transform the diagonal diffusion terms as direct finite differences 6
∑β m =1
m
⎡ ⎛ 1 ∂Cil ∂C ∂C ⎞ ⎤ * + a 2 il + a3 il ⎟ ⎥ ⎢α l ρl Dil e ⋅ ⎜ a ∂η ∂ζ ⎠ ⎦ m ⎝ ∂ξ ⎣
⎛ α ρ D* ⎞ ⎡ ∂C ⎞ ⎤ ΔV 1 ⎛ 2 ∂Cil = β1 ⎜ l l il ⎟ S1 ⎢Cil ,i +1 − Cil + + e ⋅ a3 il ⎟ ⎥ ⎜e ⋅a S1 ⎝ ∂η ∂ζ ⎠1 ⎥⎦ ⎝ ΔV ⎠1 ⎢⎣ ⎡ ⎛ α ρ D* ⎞ ΔV 2 ⎛ ∂C ⎞ ⎤ 2 ∂Cil + β 2 ⎜ l l il ⎟ S 2 ⎢Cil ,i −1 − Cil + + e ⋅ a3 il ⎟ ⎥ ⎜e ⋅a S2 ⎝ ∂η ∂ζ ⎠ 2 ⎥⎦ ⎝ ΔV ⎠2 ⎢⎣
⎡ ⎛ α ρ D* ⎞ ΔV 3 + β 3 ⎜ l l il ⎟ S3 ⎢Cil , j +1 − Cil + S3 Δ V ⎝ ⎠3 ⎣⎢
⎛ ∂C ⎞ ⎤ 1 ∂Cil + e ⋅ a3 il ⎟ ⎥ ⎜e⋅a ∂ξ ∂ζ ⎠3 ⎦⎥ ⎝
⎡ ⎛ α ρ D* ⎞ ΔV 4 + β 4 ⎜ l l il ⎟ S 4 ⎢Cil , j −1 − Cil + S4 Δ V ⎝ ⎠4 ⎢⎣
⎛ ∂C ⎞ ⎤ 1 ∂Cil + e ⋅ a3 il ⎟ ⎥ ⎜e ⋅a ∂ξ ∂ζ ⎠4 ⎦⎥ ⎝
600
13 Numerical methods for multi-phase flow in curvilinear coordinate systems
⎡ ⎛ α ρ D* ⎞ ΔV 5 + β 5 ⎜ l l il ⎟ S5 ⎢Cil ,k +1 − Cil + S5 Δ V ⎝ ⎠5 ⎢⎣
⎛ ∂C ⎞ ⎤ 1 ∂Cil + e ⋅ a 2 il ⎟ ⎥ ⎜e ⋅a ∂ξ ∂η ⎠5 ⎦⎥ ⎝
⎡ ⎛ α ρ D* ⎞ ∂C ⎞ ⎤ ΔV 6 ⎛ 1 ∂Cil + β 6 ⎜ l l il ⎟ S 6 ⎢Cil ,k −1 − Cil + + e ⋅ a 2 il ⎟ ⎥ . ⎜e ⋅a S6 ⎝ ∂ξ ∂η ⎠ 6 ⎥⎦ ⎝ ΔV ⎠ 6 ⎢⎣
(13.95)
A natural averaging of the diffusion coefficients is then the harmonic averaging as given in Appendix 12.1
(
)(
)
2 α l ρl Dil* α l ρ l Dil* DilC,m ⎛ α l ρl Dil* ⎞ m =⎜ S = S ⎟ m m ΔLh ,m ⎝ ΔV ⎠ m ΔVm α l ρl Dil* + ΔV α l ρl Dil*
(
)
(
)
,
(13.96)
m
where in the right hand side m = 1, 2, 3, 4, 5, 6 is equivalent to i + 1, i - 1, j + 1, j - 1, k + 1, k - 1, respectively regarding the properties inside a control volumes. It guaranties that if the field in one of the neighboring cells is missing the diffusion coefficient is zero. 13.4.4.2 Orthogonal coordinate systems In the case of orthogonal coordinate systems we see that:
• •
the off-diagonal diffusion terms are equal to zero, the finite volume approximations of the diagonal terms are obtained without the need to know anything about the contravariant vectors.
This illustrates the advantage of using orthogonal coordinate systems. This is valid for any diffusion terms in the conservation equations, e.g. the thermal heat diffusion terms in the energy conservation equations, the viscous diffusion terms in the momentum equations etc. 13.4.4.3 Off-diagonal diffusion terms in the general case The geometric coefficients of the off-diagonal diffusion terms can then be computed as follows
( ) = (e) ( a ) + (e) ( a ) + (e) (a ) ,
(13.97)
( ) = (e) ( a ) + (e) ( a ) + (e) (a ) ,
(13.98)
d12 = ( e ) ⋅ a 2 1
d13 = ( e ) ⋅ a3 1
11
12
31
1
23
1
1
13
32
1
33
1
1
21
22
21
2
23
22
2
21
22
31
2
23
32
2
2
= − ( d12 )i −1,
(13.99)
2
= − ( d13 )i −1 ,
(13.100)
23
2
33
2
( ) = (e) ( a ) + (e) (a ) + (e) ( a ) ,
d 31 = ( e ) ⋅ a1 3
13
22
1
( ) = (e) (a ) + (e) ( a ) + (e) ( a )
d 23 = ( e ) ⋅ a3 2
12
21
( ) = (e) (a ) + (e) ( a ) + (e) ( a )
d 22 = ( e ) ⋅ a 2 2
11
1
31
3
32
11
3
33
12
3
13
3
(13.101)
13.4 Discretization of the mass conservation equations
( ) = (e) (a ) + (e) ( a ) + (e ) (a ) ,
d 33 = ( e ) ⋅ a3 3
43
12
4
13
4
4
41
42
31
4
43
32
4
= − ( d 31 ) j −1 ,
(13.103)
= − ( d33 ) j −1,
(13.104)
33
4
4
(13.106)
( ) = (e) ( a ) + (e) ( a ) + (e) ( a )
(13.107)
51
52
11
5
51
61
53
23
5
62
11
5
22
5
6
13
5
52
21
5
53
12
5
63
12
6
5
61
6
62
21
6
63
22
6
= − ( d 51 )k −1 ,
13
6
6
( ) = (e) (a ) + (e) (a ) + (e) (a )
d 62 = ( e ) ⋅ a 2 6
42
11
( ) = (e) (a ) + (e) ( a ) + (e) ( a ) ,
d 61 = ( e ) ⋅ a1 6
41
4
(13.102)
3
(13.105)
d 52 = ( e ) ⋅ a 2 5
33
3
( ) = (e) (a ) + (e) (a ) + (e) ( a ) ,
d 51 = ( e ) ⋅ a1 5
33
32
3
( ) = (e) (a ) + (e) (a ) + (e) (a )
d 43 = ( e ) ⋅ a3 4
32
31
( ) = (e ) ( a ) + (e) ( a ) + (e) (a )
d 41 = ( e ) ⋅ a1 4
31
3
601
23
6
= − ( d 52 )k −1 . (13.108)
With this notation the diffusion term takes the form ⎡ ⎛ 1 ∂Cil ∂C ∂C ⎞ ⎤ * + a 2 il + a3 il ⎟ ⎥ ⎢α l ρl Dil e ⋅ ⎜ a ∂η ∂ζ ⎠ ⎥⎦ m m =1 ⎢⎣ ⎝ ∂ξ C 6 ⎞ D ⎛ ΔV m = ∑ β m il ,m ⎜⎜ Cil ,m − Cil + DI _ Cil ,m ⎟⎟ ΔLh ,m ⎝ Sm m =1 ⎠ 6
∑β
m
(13.109)
where DI _ Cil ,1 = d12
∂Cil ∂C + d13 il , ∂η 1 ∂ζ 1
DI _ Cil ,2 = d 22
∂Cil ∂η
+ d 23 2
∂Cil , ∂ζ 2
(13.110-111) DI _ Cil ,3 = d31
∂Cil ∂C + d33 il , ∂ξ 3 ∂ζ 3
DI _ Cil ,4 = d 41
∂Cil ∂ξ
+ d 43 4
∂Cil , ∂ζ 4
(13.112-113) DI _ Cil ,5 = d51
∂Cil ∂C + d52 il , ∂ξ 5 ∂η 5
DI _ Cil ,6 = d 61
∂Cil ∂C + d 62 il . ∂ξ 6 ∂η 6
(13.114-115) The twelve concentration derivatives are computed as follows ∂Cil 1 = ( Cil , j +1 + Cil ,i +1, j +1 − Cil , j −1 − Cil ,i+1, j −1 ) , ∂η 1 4
(13.116)
∂Cil 1 = ( Cil ,k +1 + Cil ,i +1,k +1 − Cil ,k −1 − Cil ,i +1,k −1 ), ∂ζ 1 4
(13.117)
602
13 Numerical methods for multi-phase flow in curvilinear coordinate systems
∂Cil ∂η ∂Cil ∂ζ
∂Cil ∂ξ ∂Cil ∂ζ ∂Cil ∂ξ ∂Cil ∂ζ
∂Cil ∂ξ ∂Cil ∂η ∂Cil ∂ξ ∂Cil ∂η
=
1 ( Cil ,i−1 j+1 + Cil , j+1 − Cil ,i−1 j−1 − Cil , j−1 ), 4
(13.118)
=
1 ( Cil ,k +1 + Cil ,i−1,k +1 − Cil ,k −1 − Cil ,i−1,k −1 ), 4
(13.119)
=
1 ( Cil ,i+1 + Cil ,i+1, j+1 − Cil ,i−1 − Cil ,i−1, j+1 ), 4
(13.120)
=
1 ( Cil ,k +1 + Cil , j +1,k +1 − Cil ,k −1 − Cil , j+1,k −1 ) , 4
(13.121)
=
1 ( Cil ,i+1 + Cil ,i+1, j −1 − Cil ,i−1 − Cil ,i−1, j−1 ), 4
(13.122)
=
1 ( Cil ,k +1 + Cil , j −1,k +1 − Cil ,k −1 − Cil , j −1,k −1 ), 4
(13.123)
=
1 (Cil ,i+1,k +1 + Cil ,i+1 − Cil ,i−1,k +1 − Cil ,i−1 ), 4
(13.124)
=
1 (Cil , j+1 + Cil , j+1,k +1 − Cil , j−1 − Cil , j−1,k +1 ) , 4
(13.125)
=
1 ( Cil ,i+1 + Cil ,i+1,k −1 − Cil ,i−1 − Cil ,i−1,k −1 ) , 4
(13.126)
=
1 ( Cil , j+1 + Cil , j+1,k −1 − Cil , j−1 − Cil , j−1,k −1 ) . 4
(13.127)
2
2
3
3
4
4
5
5
6
6
13.4.4.4 Final form of the finite volume concentration equation Thus the final form of the discretized concentration equation (13.1) is 6
⎛
m =1
⎝
α la ρla ( Cil − Cila ) γ va − Δτ ∑ ⎜⎜ Blm− + β m 6
= Δτγ v DCil + Δτ ∑ β m m =1
DilC,m ΔLh ,m
⎞ + ⎟⎟ Cil ,m − Cil + Δτγ v μl Cil ⎠
(
)
DilC,m ΔV m DI _ Cil ,m , ΔLh,m S m
Solving with respect to the unknown concentration we obtain
(13.128)
13.5 Discretization of the entropy equation
603
⎧ ⎡ ⎤⎫ ⎪ ⎢ ⎥⎪ ⎪ ⎢ Blm− Cil ,m ⎥⎪ 6 ⎪ ⎢ ⎥⎪ α la ρlaγ va Cila + Δτ ⎨γ v DCil + ∑ ⎢ ⎥⎬ m =1 ⎪ ⎪ C ⎢ ⎞⎥ ⎪ D ⎛ ΔV m ⎪ ⎢ + β m il ,m ⎜ Cil ,m + ⎥ DI _ C ⎟ il ,m ⎟ ⎪ ⎪ ΔLh,m ⎜⎝ Sm ⎢⎣ ⎠ ⎥⎦ ⎭ ⎩ Cil = . C 6 ⎛ ⎡ ⎤ ⎞ D il , m α laγ va ρla + Δτ ⎢γ v μ l+ + ∑ ⎜⎜ Blm− + β m ⎟⎥ ΔLh,m ⎟⎠ ⎥⎦ m =1 ⎝ ⎢⎣ (13.129) For the case of a just originating velocity field, α la = 0 and 6
⎛
m =1
⎝
γ v μl+ + ∑ ⎜⎜ blm−α l ,m ρl ,m + β m
DilC,m ⎞ ⎟ > 0, ΔLh ,m ⎟⎠
(13.130)
we have 6
6
⎡
m =1
m =1
⎢⎣
⎛ ⎞⎤ ΔV m DI _ Cil ,m ⎟⎟ ⎥ ⎜⎜ Cil ,m + Sm ⎝ ⎠ ⎥⎦ . C Dil ,m ⎞ + βm ⎟ ΔLh ,m ⎟⎠ (13.131)
γ v DCil + ∑ γ v DCil + ∑ ⎢ Blm−Cil ,m + β m Cil =
6
⎛
m =1
⎝
γ v μl+ + ∑ ⎜⎜ Blm−
DilC,m ΔLh ,m
3507"Fkuetgvk|cvkqp"qh"vjg"gpvtqr{"gswcvkqp" The entropy equation (10.62) is discretized following the procedure already described in the previous section. The result is 6
α la ρla ( sl − sla ) γ va − Δτ ∑ Blm− ( sl ,m − sl ) + Δτγ v μ l+ sl = Δτγ v Dsl* ,
(13.132)
m=1
where
⎡ 1 DlT,m ⎛ ⎞ ⎤ ΔV m DI _ Tl ,m ⎟⎟ ⎥ ⎢ ⎜⎜ Tl ,m − Tl + Sm ⎢ Tl ΔLh ,m ⎝ ⎠ ⎥ 1 6 ⎢ ⎥. * Dsl = Dsl + ∑ β m ⎢ ⎥ γ v m=1 ⎢ D sC imax ⎛ ⎞⎥ ΔV m l ,m ⎢+ ∑ ⎜ Cil ,m − Cil + S DI _ Cil ⎟⎟⎥⎥ ⎢⎣ ΔLh ,m i =2 ⎜⎝ m ⎠⎦
(13.133)
604
13 Numerical methods for multi-phase flow in curvilinear coordinate systems
The term DI _ Tl ,m is computed by replacing the concentrations in Eqs. (11.110-127) with the corresponding temperatures. The computation of the harmonic averaged thermal conductivity coefficients is given in Appendix 12.1. Solving with respect to the unknown specific entropy we obtain ⎛
sl =
6
⎞
α la ρlaγ va sla + Δτ ⎜ γ v Dsl* + ∑ Blm− slm ⎟ α la ρlaγ va
m =1 ⎝ ⎠. ⎛ 6 +⎞ + Δτ ⎜ ∑ Blm− + γ v μl ⎟ ⎝ m =1 ⎠
(13.134)
For the case of a just originating velocity field, α la = 0 and
6
∑B m =1
lm −
+ γ v μl+ > 0 ,
we have 6
sl =
γ v Dsl* + ∑ Blm− slm m =1
6
∑B m =1
lm −
+ γ vμ
.
(13.135)
+ l
3508"Fkuetgvk|cvkqp"qh"vjg"vgorgtcvwtg"gswcvkqp" The temperature equation (10.65) is discretized following the procedure already described in the previous section. The result is 6
⎛
m=1
⎝
α la ρla c pla (Tl − Tla ) γ va − Δτ ∑ ⎜⎜ Blm− c pl ,m + β m
DlT,m ΔLh ,m
⎞ ⎟⎟ (Tl ,m − Tl ) ⎠
6 ⎡ ⎤⎡ ⎤ ⎛ ∂h ⎞ B ⎢1 − ρl ⎜ l ⎟ ⎥ ⎢α la ( p − pa ) γ va − Δτ ∑ lm − ( pm − p ) ⎥ ⎢⎣ m =1 ρ lm ⎝ ∂p ⎠Tl ,all _ C ' s ⎥⎦ ⎣ ⎦ imax ⎡ ⎤ = Δτγ v ⎢ DTl N − Tl ∑ Δsilnp ( μil − Cil μl ) ⎥ i =2 ⎣ ⎦ imax 6 ⎡ DC ⎤ ⎛ ⎞ DT Δ V m ΔV m −Δτ ∑ β m ⎢ l ,m Tl ∑ Δsilnp ⎜⎜ Cil ,m − Cil + DI _ Cil ⎟⎟ − l ,m DI _ Tl ⎥ . Sm m =1 ⎢⎣ ΔLh ,m i =2 ⎥⎦ ⎝ ⎠ ΔLh ,m Sm (13.136)
13.8 Discretization of the x momentum equation
605
3509"Fkuetgvk|cvkqp"qh"vjg"rctvkeng"pwodgt" fgpukv{"gswcvkqp" The particle number density equation (10.66) is discretized following the procedure already described in the previous section. The result is 6 ⎡ m ⎤ Dn nlγ v − nlaγ va + Δτ ∑ β m ⎢( e ) ⋅ ( Vl ,m − Vcs ) nl ,m − l ,m ( nl ,m − nl ) ⎥ ΔLh ,m m =1 ⎣ ⎦ n 6 Dl ,m ΔV m = Δτγ v nl ,kin − nl ,coal + nl ,sp + Δτ ∑ β m DI _ nl ,m . ΔLh ,m S m m =1
(
)
(13.137)
The turbulent diffusion coefficient is again a result of harmonic volume averaging – Appendix 12.1. Dln,m ΔLh ,m
⎛ ν lt ⎜ t = ⎜ Sc ⎜ ΔV ⎜ ⎝
⎛ ν t ⎞⎛ ν t ⎞ ⎞ 2⎜ l t ⎟⎜ l t ⎟ ⎟ ⎝ Sc ⎠ ⎝ Sc ⎠ m . ⎟ Sm = Sm ⎛ ν lt ⎞ ⎛ ν lt ⎞ ⎟ Δ V + Δ V ⎟ ⎜ t⎟ m ⎜ t ⎟ ⎠m ⎝ Sc ⎠ ⎝ Sc ⎠ m
(13.138)
350:"Fkuetgvk|cvkqp"qh"vjg"x oqogpvwo"gswcvkqp" The x momentum equation (10.68) is discretized as already discussed in the previous Section. The result is 6
⎪⎧
m =1
⎪⎩
α la ρla ( ul − ula ) γ va − Δτ ∑ ⎨ Blm− + β m
Dlν,m ΔLh,m
⎡ 1 m1 m1 ⎤ ⎪⎫ ⎢1 + 3 ( e ) ( e ) ⎥ ⎬ ul ,m − ul ⎣ ⎦ ⎪⎭
⎛ ∂p ∂p ∂p ⎞ +Δτα l ⎜ a11γ ξ + a 21γ η + a 31γ ζ ∂ ξ ∂ η ∂ ζ ⎟⎠ ⎝ ⎡ 3 ⎤ d d ⎢ −Δτγ v ∑ cml ΔVml ( um − ul ) + cwl ΔVwl ( ucs − ul ) ⎥ ⎢ m=1 ⎥ ⎢⎣ m≠l ⎥⎦ lmax
−Δτγ v ∑ cmlL ⎣⎡( vl − vm ) bm ,3 − ( wl − wm ) bm ,2 ⎦⎤ m=1 m≠ l
−Δτγ v cwlL ⎡⎣( vl − vcs ) bw,3 − ( wl − wcs ) bw,2 ⎤⎦ 3 ⎛ ∂Δuml ∂Δuml ∂Δuml ∂Δuml ⎞ −Δτγ v ∑ cmlvm ⎜ +V1 +V 2 +V 3 ⎟ ∂τ ∂ ξ ∂ η ∂ζ ⎠ m=1 ⎝ m≠ l
(
)
606
13 Numerical methods for multi-phase flow in curvilinear coordinate systems
⎛ ∂Δucsl ∂Δucsl ∂Δucsl ∂Δucsl ⎞ −γ v Δτ cwlvm ⎜ +V1 +V 2 +V 3 ⎟ ∂ξ ∂η ∂ζ ⎠ ⎝ ∂τ 3, w ⎡ ⎤ = Δτγ v ⎢ −α l ρl g x + ∑ ⎡⎣ μ ml ( um − ul ) ⎤⎦ + μlw ( ulw − ul ) ⎥ m =1 ⎣ ⎦
Dlν,m ΔLh ,m
6
+Δτ ∑ β m m =1
2 m1 ⎛ ⎞ b uT ⎜ DI _ ul ,m − 3 ( e ) DI _ ul ,m + DI _ vislm ⎟ ⎜ ⎟ ⎜ ⎟. ⎜ 1 m1 ⎟ m2 m3 ⎜⎜ + ( e ) ⎡⎣( e ) ( vl ,m − vl ) + ( e ) ( wl ,m − wl ) ⎤⎦ ⎟⎟ ⎝ 3 ⎠
(13.139)
A natural averaging of the diffusion coefficients is the harmonic averaging – Appendix 12.1 2 (α le ρlν l* )(α le ρ lν l* ) Dilν ,m ⎛ α le ρ lν l* ⎞ m . =⎜ S = S ⎟ m m ΔLh ,m ⎝ ΔV ⎠ m ΔVm (α le ρlν l* ) + ΔV (α le ρlν l* ) m
(13.140)
It is valid for all momentum equations. Note how we arrive to the integral form of the pressure term: 1 ΔV
Δτ
∫ ∫∫∫ 0
ΔV
⎛ ∂p ∂p ∂p ⎞ gα l ⎜ a11γ ξ + a 21γ η + a31γ ζ ⎟ ∂ξ∂η∂ζ ∂τ ∂ξ ∂η ∂ζ ⎠ ⎝
⎛ ∂p ∂p ∂p ⎞ = Δτα l ⎜ a11γ ξ + a 21γ η + a 31γ ζ ⎟. ∂ξ ∂η ∂ζ ⎠ ⎝
(13.141)
The b coefficients of in the lift force expressions result from the Cartesian component decomposition: bm ,1 = a12
∂wm ∂v ∂w ∂v ∂w ∂v − a13 m + a 22 m − a 23 m + a32 m − a33 m , ∂ξ ∂ξ ∂η ∂η ∂ζ ∂ζ
(13.142)
bm ,2 = a13
∂um ∂w ∂u ∂w ∂u ∂w − a11 m + a 23 m − a 21 m + a33 m − a31 m , ∂ξ ∂ξ ∂η ∂η ∂ζ ∂ζ
(13.143)
bm ,3 = a11
∂vm ∂u ∂v ∂u ∂v ∂u − a12 m + a 21 m − a 22 m + a31 m − a32 m , ∂ξ ∂ξ ∂η ∂η ∂ζ ∂ζ
(13.144)
bw,1 = a12
∂wcs ∂v ∂w ∂v ∂w ∂v − a13 cs + a 22 cs − a 23 cs + a32 cs − a33 cs , ∂ξ ∂ξ ∂η ∂η ∂ζ ∂ζ
(13.145)
bw,2 = a13
∂ucs ∂w ∂u ∂w ∂u ∂w − a11 cs + a 23 cs − a 21 cs + a33 cs − a31 cs , ∂ξ ∂ξ ∂η ∂η ∂ζ ∂ζ
(13.146)
and
13.9 Discretization of the y momentum equation
bw,3 = a11
∂vcs ∂u ∂v ∂u ∂v ∂u − a12 cs + a 21 cs − a 22 cs + a31 cs − a32 cs . ∂ξ ∂ξ ∂η ∂η ∂ζ ∂ζ
607
(13.147)
We proceed in a similar way for the other momentum equations.
350;"Fkuetgvk|cvkqp"qh"vjg"y"oqogpvwo"gswcvkqp" The result of the discretization of the y momentum equation (10.73) is 6
⎪⎧
m=1
⎩⎪
α la ρla ( vl − vla ) γ va − Δτ ∑ ⎨ Blm− + β m
Dlν,m ΔLh ,m
m2 ⎡ 1 m2 ⎤ ⎪⎫ 1 + e e ( ) ( ) ⎢ 3 ⎥ ⎬ vl ,m − vl ⎣ ⎦ ⎭⎪
(
)
⎛ ∂p ∂p ∂p ⎞ +Δτα l ⎜ a12γ ξ + a 22γ η + a32γ ζ ⎟ ∂ξ ∂η ∂ζ ⎠ ⎝ ⎡ 3 ⎤ d d ⎢ −Δτγ v ∑ cml ΔVml ( vm − vl ) + cwl ΔVwl ( vcs − vl ) ⎥ ⎢ m=1 ⎥ ⎣⎢ m≠l ⎦⎥ lmax
−Δτγ v ∑ cmlL ⎣⎡( wl − wm ) bm ,1 − ( ul − um ) bm ,3 ⎦⎤ m=1 m≠ l
− Δτγ v cwlL ⎣⎡( wl − wcs ) bw,1 − Δulw ( ul − ucs ) bw,3 ⎦⎤ 3 ⎛ ∂Δvml ∂Δvml ∂Δvml ∂Δvml ⎞ −Δτγ v ∑ cmlvm ⎜ +V1 +V 2 +V 3 ⎟ ∂ξ ∂η ∂ζ ⎠ m=1 ⎝ ∂τ m≠ l
⎛ ∂Δvcsl ∂Δvcsl ∂Δvcsl ∂Δvcsl ⎞ −γ v Δτ cwlvm ⎜ +V1 +V 2 +V 3 ⎟ ∂ξ ∂η ∂ζ ⎠ ⎝ ∂τ 3, w ⎡ ⎤ = Δτγ v ⎢ −α l ρl g y + ∑ ⎡⎣ μ ml ( vm − vl ) ⎤⎦ + μlw ( vlw − vl ) ⎥ m =1 ⎣ ⎦
6
Dlν,m
m =1
ΔLh ,m
+Δτ ∑ β m
2 m2 ⎛ ⎞ b vT ⎜ DI _ vl ,m − 3 ( e ) DI _ ul ,m + DI _ vislm ⎟ ⎜ ⎟ ⎜ ⎟. ⎜ 1 m2 ⎟ m1 m3 ⎜⎜ + ( e ) ⎡( e ) ( ul ,m − ul ) + ( e ) ( wl ,m − wl ) ⎤ ⎟⎟ ⎣ ⎦⎠ ⎝ 3
(13.148)
608
13 Numerical methods for multi-phase flow in curvilinear coordinate systems
35032"Fkuetgvk|cvkqp"qh"vjg"z"oqogpvwo"gswcvkqp" The result of the discretization of the z momentum equation (10.77) is 6
⎪⎧
m =1
⎪⎩
α la ρla ( wl − wla ) γ va − Δτ ∑ ⎨ Blm− + β m
Dlν,m ΔLh,m
⎡ 1 m 3 m 3 ⎤ ⎪⎫ ⎢1 + 3 ( e ) ( e ) ⎥ ⎬ wl ,m − wl ⎣ ⎦ ⎪⎭
(
)
⎛ ∂p ∂p ∂p ⎞ +Δτα l ⎜ a13γ ξ + a 23γ η + a 33γ ζ ⎟ ∂ξ ∂η ∂ζ ⎠ ⎝ ⎡ 3 ⎤ −Δτγ v ⎢ ∑ cmld ΔVml ( wm − wl ) + cwld ΔVwl ( wcs − wl ) ⎥ ⎢ m=1 ⎥ ⎢⎣ m≠l ⎥⎦ lmax
−Δτγ v ∑ cmlL ⎡⎣( ul − um ) bm,2 − ( vl − vm ) bm,1 ⎤⎦ m =1 m ≠l
− Δτγ v cwlL ⎡⎣( ul − ucs ) bw,2 − ( vl − vcs ) bw,1 ⎤⎦ 3 ⎛ ∂Δwml ∂Δwml ∂Δwml ∂Δwml ⎞ −Δτγ v ∑ cmlvm ⎜ +V1 +V 2 +V 3 ⎟ ∂ξ ∂η ∂ζ ⎠ m =1 ⎝ ∂τ m ≠l
⎛ ∂Δwcsl ∂Δwcsl ∂Δwcsl ∂Δwcsl ⎞ −γ v Δτ cwlvm ⎜ +V1 +V 2 +V 3 ⎟ ∂τ ∂ ξ ∂ η ∂ζ ⎠ ⎝ 3, w ⎡ ⎤ = Δτγ v ⎢ −α l ρl g z + ∑ ⎡⎣ μ ml ( wm − wl ) ⎤⎦ + μlw ( wlw − wl ) ⎥ m =1 ⎣ ⎦ 2 m3 ⎛ ⎞ b wT ⎜ DI _ wl ,m − 3 ( e ) DI _ ul ,m + DI _ vislm ⎟ 6 ⎟ Dlν,m ⎜ +Δτ ∑ β m ⎜ ⎟. ΔLh ,m ⎜ m =1 ⎟ 1 m 3 ⎡ m1 m2 ⎜⎜ + ( e ) ( e ) ( ul ,m − ul ) + ( e ) ( vl ,m − vl ) ⎤ ⎟⎟ ⎣ ⎦ ⎝ 3 ⎠
(13.149)
35033"Rtguuwtg/xgnqekv{"eqwrnkpi" The IVA3 method: An important target of the numerical methods is to guarantee a strict mass conservation in the sense of the overall mass balance as for the single cell as well for the sum of the cells inside the physical domain of interest. We use the discretized mass conservation equations of each field in a special way to construct the so called pressure-velocity coupling keeping in mind the above requirement. First we note that the difference resulting from the time derivative divided by the new time level density can be rearranged as follows
13.11 Pressure-velocity coupling
1
(α l ρlγ v − α la ρlaγ va ) = (α l − α la ) γ v + α la (γ v − γ va ) +
ρl
609
α la ( ρl − ρla ) γ va . ρl (13.150)
Then, we divide each of the discretized field mass conservation equations by the corresponding new time level density. Having in mind Eq. (13.150) we obtain
(α l − α la ) γ v + =
1
ρl
α la 1 ( ρl − ρla ) γ va + Δτ ρl ρl
6
∑ β (ξ m
m=1
α l ρl + ξlm−α lm ρlm ) Vlmn
lm +
Δτγ v μl − α la (γ v − γ va ) .
(13.151)
We sum all of the lmax mass conservation equations. The first term disappears because the sum of all volume fractions is equal to unity. In the resulting equation the temporal density difference is replaced by the linearized form of the equation of state, Eq. (3.173),
ρl − ρla =
imax ⎛ ∂ρl ⎞ ⎛ ∂ρ ⎞ 1 p − pa ) + ⎜ l ⎟ ( sl − sla ) + ∑ ⎜⎜ 2 ( ⎟⎟ ( Ci ,l − Ci ,la ) . ala i = 2 ⎝ ∂ Ci ,l ⎠ ⎝ ∂ sl ⎠a a
(13.152) The result is lmax
( p − pa ) γ va ∑ l =1
l α la 1 + Δτ ∑ 2 ρl ala ρ l =1 l max
lmax
6
∑ β (ξ m =1
m
α l ρl + ξlm−α lm ρlm ) Vlmn = Δτ ∑ Dα l ,
lm +
l =1
(13.153) where Δτ Dα l = − γ va
α la ρl
1
ρl
Δτγ v μl − α la (γ v − γ va )
imax ⎛ ⎡⎛ ∂ρ ⎞ ∂ρl ⎢⎜ l ⎟ ( sl − sla ) + ∑ ⎜ ⎜ i = 2 ⎝ ∂ Ci ,l ⎢⎣⎝ ∂ sl ⎠ a
⎤ ⎞ ⎟⎟ ( Ci ,l − Ci ,la ) ⎥ . ⎥⎦ ⎠a
(13.154)
This equation is equivalent exactly to the sum of the discretized mass conservation equations divided by the corresponding densities. It takes into account the influence of the variation of the density with the time on the pressure change. The spatial variation of the density in the second term is still not resolved. With the next step we will derive a approximated approach to change also the influence of the spatial variation of the density on the pressure change. Writing the discretized momentum equation in the linearized form Vlmn = dVlmn − ( e ) ⋅ Vcs ,m − RVellm ( pm − p ) , m
(13.155)
610
13 Numerical methods for multi-phase flow in curvilinear coordinate systems
and replacing we finally obtain the so called pressure equation lmax
pγ va ∑ l =1
6 ⎡l ⎛ ⎤ α la ρlm ⎞ − Δ τ β ⎟ RVellm ⎥ ( pm − p ) ∑ m ⎢ ∑ ⎜ ξlm +α l + ξlm −α lm 2 ρl ala ρl ⎠ m =1 ⎣ l =1 ⎝ ⎦ max
lmax
= paγ va ∑ l =1
l α la 1 − Δ τ ∑ 2 ρl ala l =1 ρ l max
6
∑ β (ξ m=1
m
α l ρl + ξlm−α lm ρlm ) ⎡ dVlmn − ( e ) ⋅ Vcs ,m ⎤ m
lm +
lmax
+ Δτ ∑ Dα l .
⎣
⎦
(13.156)
l =1
Defining the coefficients lmax ⎛ ρ ⎞ cm = −Δτβ m ∑ ⎜ ξlm +α l + ξlm −α lm lm ⎟ RVellm , ρl ⎠ l =1 ⎝ lmax
c = pγ va ∑ l =1
6 α la − cm , ∑ ρl ala2 m=1
lmax
d = paγ va ∑ l =1
(13.157)
(13.158)
l α la + Δ τ Dα l ∑ ρl ala2 l =1 max
lmax 6 ⎛ ρ ⎞ m − Δτ ∑ β m ∑ ⎜ ξlm+α l + ξ lm−α lm lm ⎟ ⎡ dVlmn − ( e ) ⋅ Vcs ,m ⎤ , ⎣ ⎦ ρl ⎠ m =1 l =1 ⎝
(13.159)
we obtain the pressure equation 6
cp + ∑ cm pm = d ,
(13.160)
m =1
connecting each cell pressure with the pressure of the surrounding cells. We see that the system of algebraic equations has positive diagonal elements, is symmetric and strictly diagonally dominant because 6
c > ∑ cm .
(13.161)
m =1
These are very important properties. The IVA2 method: The spatial deviation of the density of the surrounding cells from the density of the cell considered can be introduced into Eq. (13.153) as follows
β m (ξlm+α l ρl + ξlm−α lm ρlm ) = β m (ξlm+α l + ξlm−α lm ) ρl + β mξ lm−α lm ( ρlm − ρ l ) . (13.161)
13.11 Pressure-velocity coupling
611
The result is lmax
( p − pa ) γ va ∑ l =1
lmax
1
l =1
ρl
+ Δτ ∑
l 6 α la + Δτ ∑∑ β m (ξlm+α l + ξlm−α lm ) Vlmn 2 ρl ala l =1 m =1 max
6
∑β m =1
ξ α lm ( ρlm − ρl ) Vlmn
m lm −
⎡ ⎢ 1 Δτγ μ − α (γ − γ ) v l la v va ⎢ ρl lmax ⎢ = ∑⎢ l =1 ⎢ imax ⎢ −γ α la ⎡⎢⎛ ∂ρl ⎞ ( s − s ) + ⎛ ∂ρl ⎜⎜ ⎜ ⎟ ∑ va l la ⎢ ρl ⎢⎣⎝ ∂ sl ⎠a i =2 ⎝ ∂ Ci ,l ⎣⎢
⎤ ⎥ ⎥ ⎥ ⎥. ⎤ ⎥⎥ ⎞ ⎟⎟ ( Ci ,l − Ci ,la ) ⎥ ⎥ ⎥⎦ ⎦⎥ ⎠a
(13.163)
The spatial density variation can again be expressed as follows, Eq. (3.173),
ρlm − ρl =
imax ⎛ ∂ρl ⎛ ∂ρ ⎞ 1 p − p ) + ⎜ l ⎟ ( slm − sl ) + ∑ ⎜⎜ 2 ( m ala i = 2 ⎝ ∂ Ci ,l ⎝ ∂ sl ⎠ a
⎞ ⎟⎟ ( Ci ,lm − Ci ,l ) . ⎠a
(13.164) With this we obtain lmax
( p − pa ) γ va ∑ l =1
l 6 ⎡ ⎛ α la p − p ⎞⎤ + Δ τ β m ⎢ξlm+α l + ξlm−α lm ⎜ 1 + m 2 ⎟ ⎥ Vlmn ∑∑ 2 ρl ala ρl ala ⎠ ⎦⎥ l =1 m =1 ⎝ ⎣⎢ max
lmax
= Δτ ∑ Dα l ,
(13.165)
l =1
where ⎡ ⎤ ⎢γ vaα la ( sl − sla ) ⎥ ⎥ ⎛ ∂ρl ⎞ ⎢ ρl Δτ Dα l = Δτγ v μl − α la ρl (γ v − γ va ) − ⎜ ⎥ ⎟ ⎢ ⎥ ⎝ ∂ sl ⎠ a ⎢ 6 ⎢ −Δτ ∑ b α ( s − s ) ⎥ lm − lm lm l ⎢⎣ ⎥⎦ m =1 imax ⎛ ∂ρl −∑ ⎜ ⎜ i = 2 ⎝ ∂ Ci ,l
6 ⎞ ⎡ ⎤ ⎟⎟ ⎢γ vaα la ( Ci ,l − Ci ,la ) − Δτ ∑ blm −α lm ( Ci ,lm − Ci ,l ) ⎥ , m =1 ⎦ ⎠a ⎣
(13.166)
Equation (13.165) is equivalent exactly to the sum of the discretized mass conservation equations. The discretized concentration equation divided by the old time level density is
612
13 Numerical methods for multi-phase flow in curvilinear coordinate systems 6
γ vaα la ( Cil − Cila ) − Δτ ∑ blm−α l ,m m=1
ρ l ,m Δτ Cil ,m − Cil ) = DCilN , ( ρla ρla
(13.167)
where
(
)
6
Dil* ,m
m =1
ΔLh ,m
DCilN = γ v DCil − μl+ Cil + ∑ β m
(C
il , m
)
− Cil + DI _ Cil ,m ,
(13.168)
does not contain time derivatives and convection terms. Even these terms are the most strongly varying in transient processes during a single time step. In the case of negligible diffusion DCilN contains only source terms. We realize that the expression on the right hand side is very similar to the left hand side of the concentration equation divided by the old time level density. A very useful approximation is then 6
γ vaα la ( Ci ,l − Ci ,la ) − Δτ ∑ blm−α lm ( Ci ,lm − Ci ,l ) m =1
(
6
)
≈ α la Cil − Cila γ va − Δτ ∑ blm−α l ,m m =1
ρl ,m Δτ Cil ,m − Cil ) = DCilN ( ρla ρla
(13.169)
and 6
γ vaα la ( sl − sla ) − Δτ ∑ blm−α lm ( slm − sl ) m =1
(
)
6
≈ α la sl − sla γ va − Δτ ∑ blm−α l ,m m =1
ρ l ,m ( sl ,m − sl ) = Δρτ DslN , ρla la
(13.170)
Thus Dα l can be approximated as follows Dα l = γ v
μl 1 ⎛ γ − γ va ⎞ − α la ⎜ v − ⎟ ρl ⎝ Δτ ⎠ ρl ρla
imax ⎛ ⎧⎪⎛ ∂ρ ⎞ ∂ρl N l ⎨⎜ ⎟ Dsl + ∑ ⎜⎜ ∂s i = 2 ⎝ ∂ Ci ,l ⎩⎪⎝ l ⎠a
⎫ ⎞ N⎪ ⎟⎟ DCil ⎬ . ⎪⎭ ⎠a (13.171)
Replacing with the normal velocities computed from the discretized momentum equation in linearized form we finally obtain the so called pressure equation lmax
pγ va ∑ l =1
l 6 ⎡ ⎛ α la p − p ⎞⎤ − Δτ ∑ β m ∑ ⎢ξ lm +α l + ξlm −α lm ⎜1 + m 2 ⎟ ⎥RVellm ( pm − p) 2 ρl ala ρl ala ⎠ ⎦⎥ m =1 l =1 ⎣ ⎢ ⎝ max
lmax
= paγ va ∑ l =1
l α la + Δτ ∑ Dα l 2 ρl ala l =1 max
lmax ⎡ ⎛ p − p ⎞⎤ m − Δτ ∑ β m ∑ ⎢ξlm+α l + ξ lm−α lm ⎜1 + m 2 ⎟ ⎥ ⎡ dVlmn − ( e ) ⋅ Vcs ,m ⎤ ⎣ ⎦ ρl ala ⎠ ⎥⎦ m =1 l =1 ⎢ ⎝ ⎣ 6
(13.172)
13.12 Staggered x momentum equation
613
or 6
cp + ∑ cm pm = d ,
(13.173)
m =1
where lmax ⎡ ⎛ p − p ⎞⎤ cm = −Δτβ m ∑ ⎢ξlm+α l + ξ lm−α lm ⎜1 + m 2 ⎟ ⎥RVellm , ρl ala ⎠ ⎦ l =1 ⎣ ⎝ lmax
c = pγ va ∑ l =1
6 α la − ∑ cm , 2 ρl ala m=1
lmax
d = paγ va ∑ l =1
(13.174)
(13.175)
l α la + Δτ ∑ Dα l 2 ρl ala l =1 max
lmax ⎡ ⎛ p − p ⎞⎤ m − Δτ ∑ β m ∑ ⎢ξlm+α l + ξ lm−α lm ⎜1 + m 2 ⎟ ⎥ ⎡ dVlmn − ( e ) ⋅ Vcs ,m ⎤ . ⎣ ⎦ ρl ala ⎠ ⎦⎥ m =1 l =1 ⎣ ⎢ ⎝ 6
(13.176)
The advantage of Eq. (13.172) for the very first outer iteration step is that it takes the influence of all sources on the pressure change which is not the case in Eq. (13.156). The advantage of Eq. (13.156) for all subsequent outer iterations is that it reduces the residuals to very low value which is not the case with Eq. (13.172) because of the approximations (13.169) and (13.170). An additional source of numerical error is that the new density is usually computed within the outer iteration by using Eq. (13.159) and not Eq. (13.164). Combined, both equations result in a useful algorithm. As a predictor step use Eq. (13.172) and for all other iterations use Eq. (13.156).
35034"Uvciigtgf"x"oqogpvwo"gswcvkqp Two families of methods are known in the literature for solving partial differential equations with low order methods, the so called co-located and staggered grid methods. In the co-located methods all dependent variables are defined at the center of the mass control volume. In these methods unless the staggered grid method is used, discretization of order higher then the first order is required to create a stable numerical method. In the staggered grid method all dependent variables are defined in the center of the mass control volume except the velocities which are defined at the faces of the volume. In both cases the velocities are required for the center as well for the faces, so that the one group of velocities is usually computed by interpolation from the known other group. The control volume for the staggered grid methods consists of the half of the volumes belonging to each face. Strictly speaking the required geometrical information that has to be stored for these methods is four times those for the co-located methods. A compromise between low order methods
614
13 Numerical methods for multi-phase flow in curvilinear coordinate systems
using low storage and stability is to derive the discretized form of the momentum equation in the staggered cell from already discretized momentum equations in the two neighboring cells. This is possible for the following reason. Momentum equations are force balances per unit mixture volume and therefore they can be volumetrically averaged over the staggered grids. In this section we will use this idea. As already mentioned the staggered computational cell in the ξ direction consists of the half of the mass control volumes belonging to the both sites of the ξ face. We will discretize the three components of the momentum equation in this staggered cell. Then we will use the dot product of the so discretized vector momentum equation with the unit face vector to obtain the normal velocity at the cell face. In doing this, we will try to keep the computational effort small by finding common coefficients for all three equations. This approach leads to a pressure gradient component normal to the face instead of a pressure gradient to each of the Cartesian directions, which is simply numerically treated. This is the key for designing implicit or semi-implicit methods.
(
)
Time derivatives: We start with the term α la ρla ul − ula γ va , perform volume averaging
α la ρla ( ul − ula ) γ va
(
= α la ρla ul − ula
) ΔVγ +ΔΔVV va
(
+ α la ,i +1 ρla ,i +1 ul ,i +1 − ula ,i +1
i +1
γ va ,i +1ΔVi +1
) ΔV + ΔV
(13.177)
i +1
and approximate the average with
α la ρla ( ul − ula ) γ va ≈ (α la ρlaγ va )u ( ulu − ulau ) ,
(13.178)
where
(α la ρlaγ va )u = α la ρla
γ va ΔV ΔV + ΔVi +1
+ α la ,i +1 ρ la ,i+1
γ va ,i +1ΔVi +1 ΔV + ΔVi +1
.
(13.179)
Note that this procedure of averaging does not give ulu =
1 ul + ul ,i +1 2
(
)
(13.180)
in the general case. Similarly we have for the other directions momentum equations
α la ρla ( vl − vla ) γ va ≈ (α la ρlaγ va )u ( vlu − vlau ) ,
(13.181)
α la ρla ( wl − wla ) γ va ≈ (α la ρlaγ va )u ( wlu − wlau ).
(13.182)
We realize that in this way of approximation the component velocity differences for all three Cartesian directions possess a common coefficient.
13.12 Staggered x momentum equation
615
Convective terms: The following approximation for the convective terms is proposed 6
α l , m ρ l , m ( ul , m − ul
∑b
lm−
m =1
=
1 ΔV + ΔVi+1 6
)
6 6 ⎧ ⎫ ⎨ΔV ∑ Blm− ul ,m − ul + ΔVi +1 ∑ ⎡⎣ Blm− ul ,m − ul ⎤⎦ ⎬ m=1 i +1 ⎭ ⎩ m=1
(
(
)
(
)
)
≈ ∑ blmu −α lu,m ρlu,m ulu,m − ulu , m =1
(13.183)
where
(
blmu −α lu,m ρlu,m = C * Blm− + 1 − C *
)(B )
lm − i +1
,
(13.184)
is the m-th face mass flow into the staggered cell divided by its volume and C* =
ΔV . ΔV + ΔVi+1
(13.185)
As in the case of the time derivatives we realize that in this way of approximation the component velocity differences for all three Cartesian directions possess a common coefficient.
Diagonal diffusion terms: We apply a similar procedure to the diagonal diffusion terms Dlν,m ΔLh ,m
6
∑ βm m =1
6
= C* ∑ β m m=1
(
+ 1 − C*
)
⎡ 1 m1 m1 ⎤ ⎢1 + 3 ( e ) ( e ) ⎥ ul ,m − ul = ⎣ ⎦
(
Dlν,m ΔLh ,m
⎡ 1 m1 m1 ⎤ ⎢1 + 3 ( e ) ( e ) ⎥ ul ,m − ul ⎣ ⎦
⎧⎪ Dlν,m β ⎨ m ∑ m=1 ⎩ ⎪ ΔLh ,m 6
)
(
)
⎡ 1 m1 m1 ⎤ ⎢1 + 3 ( e ) ( e ) ⎥ ul ,m − ul ⎣ ⎦
(
⎫
)⎪⎬
⎭⎪i+1
⎛ ⎞ Dν 1 ⎜ C * β m l ,m ⎡⎢1 + ( e )m1 ( e )m1 ⎤⎥ ⎟ ΔLh ,m ⎣ 3 ⎜ ⎟ ⎦ 6 ⎜ ⎟ u u ≈ ∑⎜ ⎟ ul ,m − ul . m =1 ⎜ ⎧ ⎫ ⎟ Dν ⎜ + ⎪⎨ 1 − C * β m l ,m ⎡1 + 1 ( e )m1 ( e )m1 ⎤ ⎪⎬ ⎟ ⎥ ⎟ ⎜ ⎪⎩ ΔLh ,m ⎢⎣ 3 ⎦ ⎪⎭i +1 ⎠ ⎝
(
(
)
)
(13.186)
616
13 Numerical methods for multi-phase flow in curvilinear coordinate systems
Thus the combined convection-diffusion terms are finally approximated as follows 6 ⎧ Dν ⎪ −∑ ⎨ Blm− + β m l ,m ΔLh ,m m =1 ⎩ ⎪ 6
(
)
⎡ 1 m1 m1 ⎤ ⎫⎪ ⎢1 + 3 ( e ) ( e ) ⎥ ⎬ ul ,m − ul ⎣ ⎦ ⎭⎪
(
6
(
)
)
≈ ∑ alm ,cd ulu,m − ulu + ∑ alm,u _ dif ulu,m − ulu , m =1
m =1
(13.187)
where ⎛ Dν alm ,cd = −C * ⎜ Blm− + β m l ,m ⎜ ΔLh ,m ⎝
⎞ * ⎟⎟ − 1 − C ⎠
(
⎛
) ⎜⎜ B
lm −
⎝
Dlν,m ΔLh ,m
+ βm
⎞ ⎟⎟ , ⎠i +1
(13.188)
⎫⎪ ⎡ Dν Dν 1 ⎧⎪ m1 m1 m1 m1 ⎤ alm ,u _ dif = − ⎨C * β m l , m ( e ) ( e ) + 1 − C * ⎢ β m l , m ( e ) ( e ) ⎥ ⎬ . 3⎪ ΔLh , m ⎣⎢ ΔLh , m ⎦⎥ i +1 ⎪⎭ ⎩ (13.189)
(
)
Similarly we have 6 ⎧ Dν ⎪ −∑ ⎨ Blm − + β m l , m ΔLh , m m =1 ⎪ ⎩ 6
(
⎡ 1 m 2 m 2 ⎤ ⎪⎫ ⎢1 + 3 ( e ) ( e ) ⎥ ⎬ vl , m − vl ⎣ ⎦ ⎪⎭
(
6
)
(
)
)
≈ ∑ alm ,cd vlu, m − vlu + ∑ alm ,v _ dif vlu, m − vlu , m =1
m =1
(13.190)
⎫⎪ ⎡ Dν Dν 1 ⎧⎪ m2 m2 m2 m2 ⎤ alm , v _ dif = − ⎨C * β m l , m ( e ) ( e ) + 1 − C * ⎢ β m l , m ( e ) ( e ) ⎥ ⎬ . 3⎪ ΔLh , m ⎢⎣ ΔLh , m ⎦⎥ i +1 ⎭⎪ ⎩ (13.191)
(
6 ⎧ Dν ⎪ −∑ ⎨ Blm − + β m l , m ΔLh , m m =1 ⎪ ⎩ 6
(
)
)
⎡ 1 m 3 m 3 ⎤ ⎫⎪ ⎢1 + 3 ( e ) ( e ) ⎥ ⎬ wl , m − wl ⎣ ⎦ ⎪⎭
(
6
(
)
)
≈ ∑ alm ,cd wlu, m − wlu + ∑ alm , w _ dif wlu, m − wlu , m =1
m =1
Dν 1 ⎧⎪ m3 m3 alm , w _ dif = − ⎨C * β m l , m ( e ) ( e ) + 1 − C * 3⎪ ΔLh , m ⎩
(
(13.192) ⎡
) ⎢β ⎢⎣
Dlν, m m
ΔLh , m
( e) ( e) m3
⎤ ⎫⎪ ⎥ ⎬. ⎥⎦ i +1 ⎭⎪ (13.193)
m3
We realize again that the coefficients alm, cd are common for all the momentum equations in the staggered cell.
13.12 Staggered x momentum equation
617
Drag force terms: The following approximation contains in fact computation of the volume averages of the linearized drag coefficients. ⎡
⎤
3
γ v ⎢ ∑ cmld ΔVml ( um − ul ) + cwld ΔVwl ( ucs − ul ) ⎥ ⎢ m =1 ⎢⎣ m ≠ l 3
⎥ ⎥⎦
(
d ≈ ∑ γ v cml m =1 m≠l
) (u u
u m
) (
− ulu + γ v cwld
) (u u
u cs
)
− ulu ,
(13.194)
where
(γ c ) d v ml
(γ c ) d v wl
u
u
(
i +1
),
(13.195)
(
i +1
).
(13.196)
=
1 γ v ΔVcmld ΔVml + γ v ,i +1ΔVi +1cmld ,i +1 ΔVml ΔV + ΔVi +1
=
1 γ v ΔVcwld ΔVcsl + γ v ,i +1ΔVi +1cwld ,i +1 ΔVcsl ΔV + ΔVi +1
Similarly we have ⎡
⎤
3
γ v ⎢ ∑ cmld ΔVml ( vm − vl ) + cwld ΔVwl ( vcs − vl ) ⎥ ⎢ m =1 ⎢⎣ m ≠ l 3
(
d ≈ ∑ γ v cml m =1 m≠l
⎥ ⎥⎦
) (v u
u m
) (
− vlu + γ v cwld
) (v
)
− vlu ,
u cs
u
(13.197)
⎡ 3 ⎤ d d ⎢ γ v ∑ cml ΔVml ( wm − wl ) + cwl ΔVwl ( wcs − wl ) ⎥ ⎢ m =1 ⎥ ⎢⎣ m ≠ l ⎥⎦ 3
(
d ≈ ∑ γ v cml m =1 m≠l
) (w u
u m
) (
d − wlu + γ v cwl
) (w u
u cs
)
− wlu .
(13.198)
Again the drag term coefficients for the momentum equations in the staggered cell are common.
Gravitational force: The volume averaging for the gravitational force gives
α l ρl g xγ v = g x (α la ρlaγ va )u ,
(13.199)
α l ρl g yγ v = g y (α la ρlaγ va )u ,
(13.200)
α l ρl g z γ v = g z (α la ρlaγ va )u .
(13.201)
618
13 Numerical methods for multi-phase flow in curvilinear coordinate systems
Interfacial momentum transfer due to mass transfer: The interfacial momentum transfer is again approximated by first volume averaging the volume mass source terms due to interfacial mass transfer. The source terms due to external injection or suction are computed exactly because it is easy to prescribe the velocities corresponding to the sources at the mass cell center. ⎡ 3, w ⎤ − ⎢ ∑ ⎡⎣ μ ml γ v ( um − ul ) ⎤⎦ + μ lwγ v ( ulw − ul ) ⎥ ⎣ m =1 ⎦ 3
≈ ∑ γ v μ ml
) (u
(γ
v
μ ml ) =
1 (γ v ΔV μ ml + γ v,i +1ΔVi +1μ ml ,i +1 ), ΔV + ΔVi +1
(13.203)
(γ
v
μ wl ) =
1 (γ v ΔV μ wl + γ v,i +1ΔVi +1μ wl ,i +1 ) , ΔV + ΔVi +1
(13.204)
(γ
v
μlw ) =
1 (γ v ΔV μlw + γ v,i +1ΔVi +1μlw,i +1 ) , ΔV + ΔVi +1
(13.205)
(γ
v
μ wl ) uwl =
1 (γ v ΔV μ wl uwl + γ v,i +1ΔVi +1μ wl ,i +1uwl ,i +1 ), ΔV + ΔVi +1
(13.206)
(γ
v
μ lw ) ulw =
1 (γ v ΔV μlwulw + γ v,i +1ΔVi +1μlw,i +1ulw,i +1 ), ΔV + ΔVi +1
(13.207)
m =1
(
u
u
u
u
u
u
u m
) (
− ulu + γ v μ wl
) (u u
u wl
) (
) (u
− ulu − γ v μ lw
u lw
u
− ulu
)
(13.202)
Similarly we have ⎡ 3, w ⎤ − ⎢ ∑ ⎡⎣ μ ml γ v ( vm − vl ) ⎤⎦ + μlwγ v ( vlw − vl ) ⎥ ⎣ m =1 ⎦ 3
(
≈ ∑ γ v μ ml m =1
) (v u
u m
) (
− vlu + γ v μ wl
) (v
) (
− vlu − γ v μlw
u wl
u
) (v
u lw
u
)
− vlu ,
(13.208)
⎡ 3, w ⎤ − ⎢ ∑ ⎡⎣ μ ml γ v ( wm − wl ) ⎤⎦ + μlwγ v ( wlw − wl ) ⎥ ⎣ m =1 ⎦ 3
(
≈ ∑ γ v μ ml m =1
) (w u
u m
) (
− wlu + γ v μ wl
) (w u
u wl
) (
− wlu − γ v μlw
) (w u
u lw
)
− wlu .
(13.209)
Lift force, off-diagonal viscous forces: The lift force and the off-diagonal viscous forces are explicitly computed by strict volume averaging.
13.12 Staggered x momentum equation
619
Pressure gradient:
α l γ ξ ( ∇p ) ⋅ i = α luγ ξ ( ∇p ) ⋅ i ,
(13.210)
where
α lu =
γ v ΔV α l + γ v ,i +1ΔVi +1α l ,i +1 . γ v ΔV + γ v ,i +1ΔVi +1
(13.211)
Similarly we have
α l γ ξ ( ∇p ) ⋅ j = α lu γ ξ ( ∇p ) ⋅ j ,
(13.212)
α l γ ξ ( ∇p ) ⋅ k = α luγ ξ ( ∇p ) ⋅k .
(13.213)
Virtual mass force: ⎛ ∂Δuml
3, cs
γ v ∑ cmlvm ⎜ m =1 m ≠l
3, cs
(
⎝ ∂τ
=∑ γ c m =1 m≠l
vm v ml
)
+V 1
∂Δuml ∂Δuml ∂Δuml ⎞ +V 2 +V 3 ⎟ ∂ξ ∂η ∂ζ ⎠
⎛ u u u u ⎜ um − ul − uma − ula + ⎜ Δτ Δτ ⎜ u ⎜ ⎜ ⎛ ⎜ + V 2 ⎜ ∂um − ∂ul 1 ⎜ ∂η ⎜ ∂η `1 ⎝ ⎝
( )
⎞ um ,i +1 − ul ,i +1 − um + ul ⎟ ⎟ 1 ⎟ ⎟ . (13.214) ⎟ ⎞ ⎛ ∂u ∂u ⎞ 3 ⎟⎟ + V 1 ⎜⎜ m − l ⎟⎟ ⎟ ⎟ `1 ⎠ ⎝ ∂ζ `1 ∂ζ `1 ⎠ ⎠
( )(
)
V1
( )
Similarly we have ⎛ ∂Δvml
3, cs
γ v ∑ cmlvm ⎜ m =1 m ≠l
3, cs
⎝ ∂τ
(
= ∑ γ v cmlvm m =1 m≠l
3, cs
∂Δvml ∂Δvml ∂Δvml ⎞ +V 2 +V 3 ⎟ ∂ξ ∂η ∂ζ ⎠
⎛ u u u u ⎜ vm − vl − vma − vla + ⎜ Δτ Δτ ⎜ u ⎜ ⎜ ⎛ ⎜ + V 2 ⎜ ∂vm − ∂vl 1 ⎜ ∂η ⎜ ∂η `1 ⎝ ⎝
( )
⎛ ∂Δwml
γ v ∑ cmlvm ⎜ m =1 m ≠l
)
+V 1
⎝ ∂τ
+V1
⎞ ⎟ − v − v + v m , i +1 l , i +1 m l ⎟ 1 ⎟ ⎟, ⎟ ⎞ ⎛ ∂vm ∂vl ⎞ ⎟ 3 + V − ⎟⎟ ⎜ ⎟ 1 ⎜ ∂ζ ∂ζ `1 ⎟⎠ ⎟⎠ `1 ⎠ `1 ⎝
(V ) ( v 1
( )
∂Δwml ∂Δwml ∂Δwml ⎞ +V 2 +V 3 ⎟ ∂ξ ∂η ∂ζ ⎠
)
(13.215)
620
13 Numerical methods for multi-phase flow in curvilinear coordinate systems
⎛ u u u u ⎜ wm − wl − wma − wla + V 1 w m , i +1 − wl , i +1 − wm + wl ⎜ Δτ 1 Δτ ⎜ u ⎜ ⎜ ⎛ ⎞ ⎛ ⎞ ⎜ + V 2 ⎜ ∂wm − ∂wl ⎟ + V 3 ⎜ ∂wm − ∂wl ⎟ ⎜ ⎟ ⎜ ⎟ 1 1 ⎜ ⎝ ∂η `1 ∂η `1 ⎠ ⎝ ∂ζ `1 ∂ζ `1 ⎠ ⎝
( )(
3, cs
(
= ∑ γ v cmlvm m =1 m≠l
)
( )
( )
)
⎞ ⎟ ⎟ ⎟ ⎟. ⎟ ⎟ ⎟ ⎠ (13.216)
Implicit treatment of the interfacial interaction: 6 ulu − ulau + ∑ alm ,cd ulu, m − ulu + α l ,u γ ξ ( ∇p ) ⋅ i Δτ m =1
(
(α la ρlaγ va )u
3, cs
(
−∑ γ v cmlvm m =1 m≠l
)
)
⎛ u u u u ⎜ um − ul − uma − ula + ⎜ Δτ Δτ ⎜ u ⎜ ⎜ ⎛ ⎜ + V 2 ⎜ ∂um − ∂ul ⎜ ∂η 1 ⎜ ∂η `1 ⎝ ⎝
(V ) ( u 1
1
(
)
− ul ,i +1 − um + ul
(
⎞ 3 ⎟⎟ + V `1 ⎠
( )
6
m , i +1
⎞ ⎟ ⎟ ⎟ L ⎟ − γ v fl ⎟ ⎟ ⎟ ⎠
⎛ ∂um ⎜ 1 ⎜ ∂ζ ⎝
( )
⎞ ⎟⎟ `1 ⎠
∂u − l ∂ζ `1
)
u
)
= −∑ alm ,u _ dif ulu, m − ulu − g x (α la ρlaγ va )u m =1
3
(
) (u
(
) (u
+ ∑ γ v μ ml m =1 3
d + ∑ γ v cml m =1 m ≠l
u
u
u m
u m
) (
) (u
) (
) (u
− ulu + γ v μ wl − ulu + γ v cwld
u
u
u wl
u cs
) (
− ulu − γ v μlw
) (u u
u lw
− ulu
)
)
− ulu + Vislu .
(13.217)
For all three velocity fields we have a system of algebraic equations with respect to the corresponding field velocity components in the x direction ⎡ 3 ⎢(α ρ γ ) 1 − a +a + γ c vm ∑ la la va u lm l ,cd v wl ⎢ Δτ m=1 m ≠l ⎣⎢
(
)
u
1 d + γ v cwl Δτ
(
) ( u
+ γ v μ wl
⎤ − γ v μ lw ⎥ ulu u⎥ ⎦⎥
) ( u
)
3
+ ∑ alm umu = bl − α l ,u γ ξ ( ∇p ) ⋅ i
(13.218)
Au uu = buu − au γ ξ ( ∇p ) ⋅ i .
(13.219)
m =1 m ≠l
or
13.12 Staggered x momentum equation
621
The non-diagonal and the diagonal elements of the Au matrix are ⎡ aml = − ⎢ γ v cmlvm ⎣
(
)
u
1 + γ v μ ml Δτ
(
) + (γ c ) d v ml
u
u
⎤ ⎥, ⎦
(13.220)
and all = (α la ρla γ va )u
3 1 − ∑ alm +al ,cd + γ v cwlvm Δτ m =1
(
)
u
m≠l
1 d + γ v cwl Δτ
(
) + (γ μ ) − (γ μ ) , v
u
wl u
v
lw u
(13.221) respectively, where 6
al , cd = −∑ alm , cd .
(13.222)
m =1
The elements of the algebraic vector buu are ⎛ uu ⎞ 3,cs bul = bul ,conv + Vislu + (α la ρlaγ va )u ⎜ la − g x ⎟ − ∑ bulm + γ v fl L ⎝ Δτ ⎠ mm =≠1l
(
⎡ + ⎢ γ v ccslvm ⎣
(
)
u
1 + γ v cwld Δτ
(
) ⎤⎥⎦ u u
u cs
(
+ ⎡ γ v μ wl ⎣
) − (γ u
v
)
u
μlw ) ⎤ ulwu u
(13.223)
⎦
where ⎡ u ⎤ u ⎢ − uma − ula + V 1 u ⎥ m , i +1 − ul , i +1 − u m + ul 1 ⎢ ⎥ Δτ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎛ ⎞ ⎛ ⎞ ∂ u ∂ u ∂ u ∂ u ⎢+ V 2 ⎜ m − l ⎟ + V 3 ⎜ m − l ⎟ ⎥ 1 ⎜ ∂η 1 ⎜ ∂ζ ⎢ ∂η `1 ⎟⎠ ∂ζ `1 ⎟⎠ ⎥⎦ `1 `1 ⎝ ⎝ ⎣ (13.224)
( )(
(
buml = −bulm = γ v cmlvm
)
u
( )
)
( )
and 6
bul ,conv = −∑ ⎡( alm ,cd + alm ,u _ dif ) ulu, m − alm ,u _ dif ulu ⎤ . ⎣ ⎦ m =1
(13.225)
The elements of the algebraic vector au are α l ,u , where l = 1,2,3. Note that by definition, if one velocity field does not exist, α l = 0 , the coefficients describing its coupling with the other fields are then equal to zero. Similarly we can discretize the momentum equations in the same staggered control volume for the other Cartesian components. The result in component form is then
622
13 Numerical methods for multi-phase flow in curvilinear coordinate systems
Au uu = buu − au γ ξ ( ∇p ) ⋅ i ,
(13.226)
Au vu = bv u − au γ ξ ( ∇p ) ⋅ j ,
(13.227)
Au w u = bw u − au γ ξ ( ∇p ) ⋅ k .
(13.228)
It is remarkable that the A matrix and the coefficients of the pressure gradient are common for all the three systems of equations. If we take the dot product of each u-v-w equation with the unit normal vector at the control volume face we then obtain
( )
( )
Au ⎡ e11 uu + e12 1 ⎣
( )
1
( )
vu + e13 w u ⎤ 1 ⎦
( )
( )
( )
= e11 bu u + e12 bv u + e13 bw u − auγ ξ ⎡ e1 ⋅ ( ∇p ) ⎤ . 1 1 1 1 ⎣ ⎦
(13.229)
Having in mind that the outwards pointing normal face velocity is V = ( e ) ⋅ Vu, n
1
(13.230)
and ∂p 1 = ( e ) ⋅ ( ∇p ) , ∂ξ
(13.231)
we obtain finally n
A u V = b u − au γ ξ
∂p , ∂ξ
(13.232)
where
( )
( )
( )
bu = e11 bu u + e12 bv u + e13 bw u . 1
1
1
(13.233)
This algebraic system can be solved with respect to each field velocity provided that det Au = a11a22 a33 + a12 a23 a31 + a21a32 a13 − a31a22 a13 − a32 a23 a11 − a12 a21a33 ≠ 0 . (13.234)
The result is V = dVξ − RVξ γ ξ ( pi +1 − p ) , n
(13.235)
Appendix 13.1 Harmonic averaged diffusion coefficients
623
where
( )
dVξ = A u
−1
bu
(13.236)
with components ⎛ 3 ⎞ dV ξ ,l = ⎜ ∑ bm alm ⎟ / det A u , ⎝ m =1 ⎠
(13.237)
and
( )
RVξ = A u
−1
au ,
(13.238)
with components ⎛ 3 ⎞ RU l = ⎜ ∑ α ma ,u alm ⎟ / det A u , ⎝ m =1 ⎠
(13.239)
and the a values are a11 = a22 a33 − a32 a23 , a12 = a32 a13 − a12 a33 , a13 = a12 a23 − a22 a13 , a21 = a23 a31 − a21a33 , a22 = a11 a33 − a31 a13 , a23 = a21a13 − a23 a11, a31 = a21a32 − a31a22 , a32 = a12 a31 − a32 a11 , a33 = a11a22 − a21a12 . (13.240-248)
Actually, not the absolute but the relative normal face velocity is required to construct the pressure equation which is readily obtained V n = V − ( e ) ⋅ Vcs. n
1
(13.249)
Crrgpfkz"3503"Jctoqpke"cxgtcigf"fkhhwukqp"eqghhkekgpvu" A natural averaging of the coefficients describing diffusion across the face m, having surface cross section S m is then the harmonic averaging DlΦ, m ΔLh , m
2 ( Φ l )( Φ l )m ⎛Φ ⎞ = ⎜ l ⎟ Sm = Sm ΔVm ( Φ l ) + ΔV ( Φ l )m ⎝ ΔV ⎠ m
where on the right hand side m = 1, 2, 3, 4, 5, 6 is equivalent to i + 1, i - 1, j + 1, j 1, k + 1, k - 1, respectively regarding the properties inside a control volumes. ΔV is the non-staggered cell volume, and ΔVm is the volume of the cell at the other side of face m. It guaranties that if the field in one of the neighboring cells is missing the diffusion coefficient is zero. This property is derived from the solution of the steady state one-dimensional diffusion equations.
624
13 Numerical methods for multi-phase flow in curvilinear coordinate systems
For computation of Dil*,m
ΔLh ,m
(
)(
)
2 α l ρl Dil* α l ρl Dil* ⎛ α ρ D* ⎞ m = ⎜ l l il ⎟ S m = S m * * Δ V Δ V α ρ D + Δ V α ρ ⎝ ⎠m m l l il l l Dil
(
)
(
)
m
we simply set Φ l = α l ρl Dil* . For computation of For computation of DilC, m ΔLh , m
DlT, m
ΔLh , m DilsC, m
ΔLh , m
we simply set Φ l = α l λl . we simply set Φ l = α l ρl Dil* ( sil − s1l ) . Note that
=0
for sil = s1l or sil , m = s1l , m . For computation of the turbulent particle diffusion coefficient ply set Φ l =
ν lt Sct
Dln,m
ΔLh, m
we sim-
.
For computation of
Dlν, m
ΔLh , m
we simply set Φ l = α l ρlν l* .
In the case of cylindrical or Cartesian coordinate systems we have zero offdiagonal diffusion terms and 2 ( Φ l )( Φ l )i +1 DlΦ1 = , Δrh Δr ( Φ l )i +1 + Δri +1Φ l 2 ( Φ l )( Φ l )i −1 DlΦ2 = , Δrh ,i −1 Δr ( Φ l )i −1 + Δri −1Φ l 2 ( Φ l )( Φ l ) j +1 DlΦ3 = , κ κ r Δθ h r ⎡ Δθ ( Φ l ) + Δθ j +1Φ l ⎤ j + 1 ⎣ ⎦ 2 ( Φ l )( Φ l ) j −1 DlΦ4 = , κ κ r Δθ h , j −1 r ⎡ Δθ ( Φ l ) + Δθ j −1Φ l ⎤ j − 1 ⎣ ⎦ 2 ( Φ l )( Φ l )k +1 DlΦ5 = , Δzh Δz ( Φ l )k +1 + Δzk +1Φ l
Appendix 13.2 Off-diagonal viscous diffusion terms of the x momentum equation
625
2 ( Φ l )( Φ l )k −1 DlΦ6 = . Δzh , k −1 Δz ( Φ l )k −1 + Δzk −1Φ l
Crrgpfkz"3504"Qhh/fkciqpcn"xkueqwu"fkhhwukqp"vgtou"qh"vjg" x"oqogpvwo"gswcvkqp" The off-diagonal viscous diffusion terms in the x momentum equation are 6
∑β m =1
m
Dlν, m ⎛ 2 m1 uT ⎞ DI _ ul , m − ( e ) DI _ ulb, m + DI _ vislm ⎟ ΔLh , m ⎜⎝ 3 ⎠
⎧⎪ ∂u ∂u = β1* ⎨qx1,12 l + qx1,13 l ∂ η ∂ζ 1 ⎩⎪ ⎧⎪ ∂u + β 2* ⎨qx 2,12 l ∂η ⎩⎪ ⎧⎪ ∂u + β 3* ⎨ qx 3,11 l ∂ξ ⎩⎪ ⎧⎪ ∂u + β 4* ⎨ qx 4,11 l ∂ξ ⎩⎪ ⎧⎪ ∂u + β 5* ⎨ qx 5,11 l ∂ξ ⎪⎩ ∂u ⎪⎧ + β 6* ⎨ qx 6,11 l ∂ξ ⎪⎩
+ qx 2,13 2
+ qx 3,13 3
+ qx 4,13 4
+ qx 5,12 5
+ qx 6,12 6
ul ∂ζ ∂ul ∂ζ ∂ul ∂ζ ∂ul ∂η ∂ul ∂η
+ qx1,22 1
qx 2,22 2
∂vl ∂v + qx 2,23 l ∂η 2 ∂ζ
+ qx 3,21 3
+ qx 4,21 4
+ qx 5,21 5
+ qx 6,21 6
∂vl ∂v ∂w ∂w ⎫⎪ + qx1,23 l + qx1,32 l + qx1,33 l ⎬ ∂η 1 ∂ζ 1 ∂η 1 ∂ζ 1 ⎪⎭
∂vl ∂ξ ∂vl ∂ξ ∂vl ∂ξ ∂vl ∂ξ
+ qx 3,23 3
+ qx 4,23 4
+ qx 5,22 5
+ qx 6,22 6
+ qx 2,32 2
∂vl ∂ζ ∂vl ∂ζ
∂wl ∂η
+ qx 3,31 3
2
∂wl ∂ξ
+ qx 4,31 4
+ qx 2,33
∂wl ∂ζ
+ qx 3,33 3
∂wl ∂ξ
∂vl ∂w + qx 5,31 l ∂η 5 ∂ξ ∂vl ∂w + qx 6,31 l ∂η 6 ∂ξ
⎫⎪ ⎬ ⎪ 2⎭
∂wl ∂ζ
⎫⎪ ⎬ ⎪ 3⎭
+ qx 4,33
∂wl ∂ζ
⎫⎪ ⎬ ⎪ 4⎭
+ qx 5,32
∂wl ∂η
⎫⎪ ⎬ 5⎪ ⎭
+ qx 6,32
∂wl ∂η
⎪⎫ . ⎬ 6⎪ ⎭
4
5
6
Here the coefficients
β m* = β m =
Dlν, m ΔV m ΔLh ,m Sm
are used also in the other momentum equations. The following 36 coefficients are functions of the geometry only. 4 11 21 1 11 12 13 ( e ) a 1 + ( e ) a 22 1 + ( e ) a 23 1 = d12 + ( e ) a 21 1, 3 3 4 11 1 11 12 13 = ( e ) a31 + ( e ) a32 + ( e ) a33 = d13 + ( e ) a31 , 1 1 1 1 3 3
qx1,12 =
( )
( )
( )
( )
qx1,13
( )
( )
( )
( )
626
13 Numerical methods for multi-phase flow in curvilinear coordinate systems
qx1,22 = ( e )
(a )
−
2 11 22 ( e) a 1 , 3
qx1,23 = ( e )
(a )
−
2 11 32 ( e ) a 1, 3
qx1,32 = ( e )
(a )
−
2 11 23 ( e ) a 1, 3
qx1,33 = ( e )
(a )
−
2 11 33 ( e ) a 1, 3
12
21
1
12
31
1
13
1
13
qx 2,12 =
21
31
1
4 21 21 (e) a 3
( )
( )
( )
( )
( ) + ( e) ( a ) + ( e) ( a ) 22
23
22
2
23
2
2
= d 22 +
1 21 21 (e) a 3
= d 23 +
1 21 31 (e) a 3
( )
2
= − ( q x1,12 )i −1 , qx 2,13 =
4 21 31 (e) a 3
( ) + ( e) ( a ) + ( e) ( a ) 22
23
32
2
33
2
2
( )
2
= − ( q x1,13 )i−1 , qx 2,22 = ( e )
22
qx 2,23 = ( e )
22
qx 2,32 = ( e )
23
qx 2,33 = ( e )
23
(a ) 21
2
(a )
2
(a )
2
31
21
(a ) 31
2
qx 3,11 =
4 31 11 (e) a 3
qx 3,13 =
4 31 31 ( e) a 3
−
2 21 22 (e) a 3
−
2 21 32 (e) a 3
−
2 21 23 (e) a 3
−
2 21 33 ( e) a 3
( )
2
= − ( qx1,22 )i −1,
2
= − ( qx1,23 )i −1,
2
= − ( qx1,32 )i −1,
2
= − ( qx1,33 )i −1,
( ) ( )
( )
( ) + (e) ( a ) + (e) ( a ) 33
12
13
3
3
( ) + (e) ( a ) + (e) ( a )
qx 3,21 = ( e )
32
qx 3,23 = ( e ) qx 3,31 = ( e )
32
3
32
33
(a ) 11
3
(a ) 31
(a )
3
11
3
32
33
32
3
3
−
2 31 12 ( e ) a 3, 3
−
2 31 32 ( e ) a 3, 3
−
2 31 13 (e) a 3, 3
( )
( )
( )
33
3
= d31 +
1 31 11 ( e ) a 3, 3
= d33 +
1 31 31 ( e ) a 3, 3
( )
( )
Appendix 13.2 Off-diagonal viscous diffusion terms of the x momentum equation
qx 3,33 = ( e )
qx 4,11 =
33
(a ) 31
3
4 41 11 (e) a 3
−
2 31 33 (e) a 3 , 3
( )
( ) + ( e) ( a ) + ( e) ( a ) 42
43
12
4
13
4
4
= d 41 +
1 41 11 (e) a 3
= d 43 +
1 41 31 (e) a 3
( )
4
= − ( qx 3,11 ) j −1 , qx 4,13 =
4 41 31 (e) a 3
( ) + (e) ( a ) + ( e) ( a ) 42
43
32
4
33
4
4
( )
4
= − ( qx 3,13 ) j −1 , qx 4,21 = ( e )
42
qx 4,23 = ( e )
42
qx 4,31 = ( e )
43
qx 4,33 = ( e )
43
(a ) 11
4
(a ) 31
(a )
4
11
4
(a ) 31
4
qx 5,11 =
4 51 11 (e) a 3
qx 5,12 =
4 51 21 ( e) a 3
2 41 12 (e) a 3
−
2 41 32 (e) a 3
−
2 41 13 (e) a 3
−
2 41 33 (e) a 3
( )
( )
( )
= − ( qx 3,21 ) j −1,
4
4
( )
= − ( qx 3,23 ) j −1 ,
4
= − ( qx 3,31 ) j −1 ,
4
= − ( qx 3,33 ) j −1 ,
( ) + (e) ( a ) + (e) ( a ) 52
53
12
5
13
5
5
( ) + (e) ( a ) + (e) ( a )
qx 5,21 = ( e )
52
(a ) 11
5
qx 5,22 = ( e )
(a )
qx 5,31 = ( e )
(a )
52
53
qx 5,32 = ( e )
qx 6,11 =
−
53
21
5
11
5
(a ) 21
5
4 61 11 (e) a 3
= − ( qx 5,11 )k −1 ,
52
53
22
5
5
−
2 51 12 ( e ) a 5, 3
−
2 51 22 ( e ) a 5, 3
−
2 51 13 ( e ) a 5, 3
−
2 51 23 (e) a 5, 3
23
1 51 11 (e) a 5 , 3
( )
= d52 +
1 51 21 ( e) a 5, 3
= d 61 +
1 61 11 ( e) a 3
( )
( )
( )
( )
( )
( ) + ( e) ( a ) + ( e) ( a ) 62
6
5
= d51 +
63
12
6
13
6
( )
6
627
628
13 Numerical methods for multi-phase flow in curvilinear coordinate systems
qx 6,12 =
4 61 21 ( e) a 3
( ) + (e) ( a ) + (e) ( a ) 62
63
22
6
23
6
6
= d62 +
1 61 21 ( e) a 3
( )
6
= − ( q x 5,12 )k −1, qx 6,21 = ( e )
qx 6,22 = ( e ) qx 6,31 = ( e )
(a )
62
11
62
qx 6,32 = ( e )
(a ) 21
(a )
63
6
6
11
63
6
(a ) 21
6
−
2 61 12 (e) a 3
−
2 61 22 (e) a 3
−
2 61 13 (e) a 3
−
2 61 23 ( e) a 3
( )
6
( )
( )
6
6
( )
6
= − ( qx 5,21 ) k −1 , = − ( qx 5,22 )k −1 , = − ( qx 5,31 )k −1, = − ( qx5,32 ) k −1 .
Crrgpfkz"3505"Qhh/fkciqpcn"xkueqwu"fkhhwukqp"vgtou"qh"vjg" y"oqogpvwo"gswcvkqp" The off-diagonal viscous diffusion terms in the y momentum equation 6
∑ βm m =1
Dlν, m ⎛ 2 m2 b vT ⎞ ⎜ DI _ vl , m − ( e ) DI _ ul , m + DI _ vislm ⎟ ΔLh , m ⎝ 3 ⎠
are computed using the same procedure as those for x equation replacing simply the subscript x with y and using the following geometry coefficients. q y1,12 = ( e )
(a )
−
2 12 21 ( e ) a 1, 3
q y1,13 = ( e )
(a )
−
2 12 31 (e) a 1, 3
11
22
1
11
32
1
( )
( )
q y1,22 = ( e )
(a )
+
4 12 22 ( e) a 3
q y1,23 = ( e )
(a )
+
4 12 32 ( e) a 3
q y1,32 = ( e )
(a )
−
2 12 23 ( e ) a 1, 3
q y1,33 = ( e )
(a )
−
2 12 33 ( e ) a 1, 3
11
11
13
13
21
1
31
1
22
1
32
1
( ) + (e) ( a ) , 13
23
1
1
( ) + (e) ( a ) , 13
1
( )
( )
33
1
Appendix 13.3 Off-diagonal viscous diffusion terms of the y momentum equation
q y 2,12 = ( e )
21
q y 2,13 = ( e )
21
q y 2,22 = ( e )
21
q y 2,23 = ( e )
21
q y 2,32 = ( e )
23
q y 2,33 = ( e )
23
q y 3,11 = ( e )
31
q y 3,13 = ( e )
31
q y 3,21 = ( e )
31
q y 3,23 = ( e ) q y 3,31 = ( e )
31
33
q y 3,33 = ( e )
33
q y 4,11 = ( e )
41
q y 4,13 = ( e )
41
q y 4,21 = ( e )
41
q y 4,23 = ( e )
41
(a ) 22
(a )
2
32
2
(a )
2
(a )
2
21
31
(a ) 22
(a )
2
32
(a )
2
12
3
(a ) 32
(a )
3
11
3
(a ) 31
(a )
3
12
3
(a ) 32
(a )
3
12
4
(a ) 32
(a )
4
11
4
(a ) 31
4
−
2 22 21 (e) a 3
−
2 22 31 (e) a 3
+
4 22 22 ( e) a 3
+
4 22 32 (e) a 3
−
2 22 23 (e) a 3
−
2 22 33 (e) a 3
−
2 32 11 (e) a 3 , 3
−
2 32 31 ( e) a 3 , 3
+
4 32 12 (e) a 3
+
4 32 32 ( e) a 3
−
2 32 13 ( e ) a 3, 3
−
2 32 33 ( e ) a 3, 3
−
2 42 11 (e) a 3
−
2 42 31 (e) a 3
+
4 42 12 (e) a 3
+
4 42 32 (e) a 3
( )
( )
2
= − q y1,12
(
)
i −1
(
)
,
= − q y1,13
2
i −1
( ) + (e) ( a ) 23
( ) + (e) ( a ) 23
( )
= − q y1,32
(
)
(
)
= − q y1,33
2
(
)
i −1
(
)
i −1
,
= − q y 3,21
(
)
j −1
(
)
2
= − q y1,23
33
2
2
= − q y1,22
23
2
( )
,
2
i −1
,
, ,
i −1
( )
( )
( ) + (e) ( a ) , 33
13
3
3
( ) + (e) ( a ) , 33
33
3
3
( )
( )
( )
4
( )
4
= − q y 3,11
(
)
(
)
= − q y 3,13
j −1
,
j −1
( ) + (e) ( a ) 43
13
4
4
( ) + (e) ( a ) 43
4
,
33
4
= − q y 3,23
, ,
j −1
629
630
13 Numerical methods for multi-phase flow in curvilinear coordinate systems
q y 4,31 = ( e )
43
q y 4,33 = ( e )
43
q y 5,11 = ( e )
51
q y 5,12 = ( e )
51
q y 5,21 = ( e )
51
q y 5,22 = ( e ) q y 5,31 = ( e )
53
q y 5,32 = ( e ) q y 6,11 = ( e )
51
53
61
q y 6,12 = ( e )
61
q y 6,21 = ( e )
61
q y 6,22 = ( e ) q y 6,31 = ( e )
61
63
q y 6,32 = ( e )
63
(a ) 12
4
(a ) 32
(a )
4
12
5
(a ) 22
(a )
5
11
5
(a ) 21
(a )
5
12
5
(a ) 22
(a )
5
12
6
(a ) 22
(a )
6
11
6
(a ) 21
(a )
6
12
6
(a ) 22
6
−
2 42 13 (e) a 3
−
2 42 33 (e) a 3
−
2 52 11 ( e ) a 5, 3
−
2 52 21 ( e) a 5, 3
+
4 52 12 (e) a 3
+
4 52 22 (e) a 3
−
2 52 13 ( e) a 5 , 3
−
2 52 23 ( e) a 5 , 3
−
2 62 11 (e) a 3
−
2 62 21 (e) a 3
+
4 62 12 ( e) a 3
+
4 62 22 ( e) a 3
−
2 62 13 (e) a 3
−
2 62 23 ( e) a 3
( )
4
( )
= − q y 3,31
(
)
(
)
= − q y 3,33
4
,
j −1
j −1
,
( )
( )
( ) + ( e) ( a ) , 53
13
5
5
( ) + ( e) ( a ) , 53
23
5
5
( )
( )
( )
6
( )
6
= − q y 5,11
(
)
(
)
= − q y 5,12
k −1
k −1
( ) + (e) ( a ) 63
13
6
6
( ) + (e) ( a )
( )
63
( )
6
, = − q y 5,21
(
)
(
)
= − q y 5,22
23
6
6
,
= − q y 5,31
(
)
(
)
= − q y 5,32
6
k −1
k −1
,
k −1
,
,
k −1
.
Crrgpfkz"3506"Qhh/fkciqpcn"xkueqwu"fkhhwukqp"vgtou"qh"vjg" z oqogpvwo"gswcvkqp" The off-diagonal viscous diffusion terms in the z momentum equation
Appendix 13.4 Off-diagonal viscous diffusion terms of the z momentum equation 6
∑ βm m =1
631
Dlν, m ⎛ 2 m3 ⎞ DI _ wl , m − ( e ) DI _ ulb,m + DI _ vislmwT ⎟ ⎜ ΔLh , m ⎝ 3 ⎠
are computed using the same procedure as those for x equation replacing simply the subscript x with z and using the following geometry coefficients. qz1,12 = ( e )
(a )
−
2 13 21 ( e ) a 1, 3
qz1,13 = ( e )
(a )
−
2 13 31 ( e ) a 1, 3
11
23
1
11
33
1
( )
( )
qz1,22 = ( e )
(a )
−
2 13 22 ( e ) a 1, 3
qz1,23 = ( e )
(a )
−
2 13 32 ( e ) a 1, 3
qz1,32 = ( e )
( a ) + (e) ( a )
+
4 13 23 (e) a 1 , 3
qz1,33 = ( e )
( a ) + (e) ( a )
+
4 13 33 (e) a 1, 3
12
23
1
12
33
1
11
( )
12
21
22
1
11
1
12
31
32
1
qz 2,12 = ( e )
21
qz 2,13 = ( e )
21
qz 2,22 = ( e )
22
qz 2,23 = ( e )
22
qz 2,32 = ( e )
21
qz 2,33 = ( e )
21
qz 3,11 = ( e )
( )
23
(a )
2
33
2
(a ) 23
(a )
2
33
2
−
2 23 21 (e) a 3
−
2 23 31 (e) a 3
−
2 23 22 (e) a 3
−
2 23 32 ( e) a 3
( )
31
(a )
22
13
3
(a ) 33
3
2
= − ( qz1,13 )i −1,
2
= − ( qz1,22 )i −1,
2
= − ( qz1,23 )i −1, 4 23 23 (e) a 3
+
4 23 33 (e) a 3
2
32
2
+
22
( a ) + ( e) ( a ) 2
= − ( qz1,12 )i −1,
( )
22
31
2
( )
( a ) + (e) ( a ) 21
( )
( )
2
31
qz 3,13 = ( e )
(a )
1
( )
−
2 33 11 ( e) a 3 , 3
−
2 33 31 ( e ) a 3, 3
( )
( )
( )
( )
2
= − ( qz1,32 )i −1,
2
= − ( qz1,33 )i −1 ,
632
13 Numerical methods for multi-phase flow in curvilinear coordinate systems
qz 3,21 = ( e )
32
qz 3,23 = ( e ) qz 3,31 = ( e )
32
31
31
41
qz 4,13 = ( e )
41
qz 4,21 = ( e )
42
qz 4,23 = ( e ) qz 4,31 = ( e )
42
41
(a ) 33
3
−
2 33 12 ( e ) a 3, 3
−
2 33 32 ( e) a 3, 3
( )
( )
( a ) + (e) ( a ) 32
11
12
41
( a ) + ( e) ( a ) 32
31
51
52
qz 5,22 = ( e )
52
51
qz 5,32 = ( e )
(a ) 13
4
(a ) 33
4
(a ) 13
4
(a ) 33
4
−
2 43 31 ( e) a 3
−
2 43 12 ( e) a 3
−
2 43 32 ( e) a 3
42
31
5
(a )
5
23
13
(a ) 23
5
4 43 13 ( e) a 3
+
4 43 33 (e) a 3
4
( )
4
( )
−
2 53 21 ( e) a 5 , 3
−
2 53 12 ( e) a 5 , 3
−
2 53 22 ( e ) a 5, 3
= − ( qz 3,31 ) j −1 ,
4
( )
( ) ( )
( )
12
5
( a ) + (e) ( a ) 52
5
= − ( qz 3,23 ) j −1 ,
4
2 53 11 ( e) a 5 , 3
52
21
= − ( qz 3,21 ) j −1 ,
−
5
51
= − ( qz 3,13 ) j −1 ,
+
32
( a ) + ( e) ( a ) 11
= − ( qz 3,11 ) j −1,
4
( ) 4
( )
4
( )
12
( )
4
( )
( a ) + ( e) ( a )
(a )
4 33 33 ( e) a 3, 3
( )
42
5
+
3
2 43 11 (e) a 3
11
13
4 33 13 (e) a 3 , 3
−
( a ) + ( e) ( a )
(a )
+
32
4
51
qz 5,21 = ( e )
3
4
qz 4,33 = ( e )
qz 5,31 = ( e )
3
3
qz 4,11 = ( e )
qz 5,12 = ( e )
13
3
qz 3,33 = ( e )
qz 5,11 = ( e )
(a )
22
5
+
4 53 13 ( e) a 5 , 3
+
4 53 23 ( e) a 5, 3
( )
( )
= − ( qz 3,33 ) j −1 ,
References
qz 6,11 = ( e )
61
qz 6,12 = ( e )
61
qz 6,21 = ( e )
62
qz 6,22 = ( e ) qz 6,31 = ( e )
62
61
qz 6,32 = ( e )
(a ) 13
6
(a )
6
(a )
6
23
13
(a ) 23
6
−
2 63 11 (e) a 3
−
2 63 21 ( e) a 3
−
2 63 12 (e) a 3
−
2 63 22 (e) a 3
( )
( )
62
11
12
6
61
6
( a ) + (e) ( a ) 62
21
6
6
= − ( qz 5,12 )k −1,
6
= − ( qz 5,21 ) k −1,
( )
( a ) + (e) ( a ) 22
6
= − ( qz 5,11 )k −1,
6
( )
633
6
= − ( qz 5,22 )k −1,
+
4 63 13 (e) a 3
+
4 63 23 ( e) a 3
( )
6
( )
6
= − ( qz 5,31 )k −1 , = − ( qz 5,32 )k −1.
Tghgtgpegu" Antal SP et al (2000) Development of a next generation computer code for the prediction of multi-component multiphase flows, Int. Meeting on Trends in Numerical and Physical Modeling for Industrial Multiphase Flow, Cargese, France Brackbill JU, Kothe DB and Zeinach C (1992) A continuum method for modelling surface tension, Journal of Computational Physics, vol 100, p 335 Gregor C, Petelin S and Tiselj I (2000) Upgrade of the VOF method for the simulation of the dispersed flow, Proc. of ASME 2000 Fluids Engineering Division Summer Meeting, Boston, Massachusetts, June 11-15 Harlow FH and Amsden AA (1975) Numerical calculation of multiphase flow, Journal of Computational Physics vol 17 pp 19-52 Hirt CW (1993) Volume-fraction techniques: powerful tools for wind engineering, J. Wind Engineering and Industrial Aerodynamics, vol 46-47, p 327 Hou S, Zou Q, Chen S, Doolen G and Cogley AC (1995) Simulation of cavity flow by the lattice Boltzmann method, J. Comput. Phys., vol 118, p 329 Jamet D, Lebaigue O, Courtis N and Delhaye (2001) The second gradient method for the direct numerical simulation of liquid-vapor flows with phase change, Jornal of Comp. Physics vol 169 pp 624-651 Kolev NI (1994) IVA4: Modeling of mass conservation in multi-phase multi-component flows in heterogeneous porous media. Kerntechnik, vol 59 no 4-5 pp 226-237 Kolev NI (1994) The code IVA4: Modelling of momentum conservation in multi-phase multi-component flows in heterogeneous porous media, Kerntechnik, vol 59 no 6 pp 249-258 Kolev NI (1995) The code IVA4: Second law of thermodynamics for multi phase flows in heterogeneous porous media, Kerntechnik, vol 60 no 1, pp 1-39 Kolev NI (1997) Comments on the entropy concept, Kerntechnik, vol 62 no 1 pp 67-70
634
13 Numerical methods for multi-phase flow in curvilinear coordinate systems
Kolev NI (1998) On the variety of notation of the energy conservation principle for single phase flow, Kerntechnik, vol 63 no 3 pp 145-156 Kolev NI (2001) Conservation equations for multi-phase multi-component multi-velocity fields in general curvilinear coordinate systems, Keynote lecture, Proceedings of ASME FEDSM’01, ASME 2001 Fluids Engineering Division Summer Meeting, New Orleans, Louisiana, May 29 - June 1, 2001 Kothe DB, Rider WJ, Mosso SJ, Brock JS and Hochstein JI (1996) Volume tracking of Interfaces having surface tension in two and three dimensions. AIAA 96-0859 Kumbaro A, Toumi I and Seignole V (April 14-18, 2002) Numerical modeling of twophase flows using advanced two fluid system, Proc. of ICONE10, 10th Int. Conf. On Nuclear Engineering, Arlington, VA, USA Lahey RT Jr and Drew D (October 3-8, 1999) The analysis of two-phase flow and heat transfer using a multidimensional, four field, two-fluid model, Ninth Int. Top. Meeting on Nuclear Reactor Thermal-Hydraulics (NURETH-9), San Francisco, California Miettinen J and Schmidt H (April 14-18, 2002) CFD analyses for water-air flow with the Euler-Euler two-phase model in the FLUENT4 CFD code, Proc. of ICONE10, 10th Int. Conf. On Nuclear Engineering, Arlington, VA, USA Nakamura T, Tanaka R, Yabe T and Takizawa K (2001) Exactly conservative semiLagrangian scheme for multi-dimensional hyperbolic equations with directional splitting technique, J. Comput. Phys., vol 147 p 171 Nourgaliev RR, Dinh TN and Sehgal BR (2002) On lattice Boltzmann modelling of phase transition in isothermal non-ideal fluid, Nuclear Engineering and Design, vol 211 pp 153-171 Osher S and Fredkiw R (2003) Level set methods and dynamic implicit surfaces, SpringerVerlag New York, Inc. Rider WJ and Kothe DB (1998) Reconstructing volume tracking. Journal of Computational Physics, vol 141 p 112 Staedke H, Franchello G and Worth B (June 8-12, 1998) Towards a high resolution numerical simulation of transient two-phase flow, Third International Conference on Multi-Phase, ICMF’98 Sussman M, Smereka P and Oslier S ( 1994) A level set approach for computing solutions to incompressible two-phase flow, Journal of Computational Physics, vol 114 p 146 Swthian JA (1996) Level set methods, Cambridge University Press Takewaki H, Nishiguchi A and Yabe T, (1985) The Cubic-Interpolated Pseudo Particle (CIP) Method for Solving Hyperbolic-Type Equations, J. Comput. Phys., vol 61 p 261 Takewaki H and Yabe Y (1987) Cubic-Interpolated Pseudo Particle (CIP) Method Application to Nonlinear Multi-Dimensional Problems, J. Cornput. Phys., vol 70 p 355 Tomiyama A et al. (2000) (N+2)-Field modelling of dispersed multiphase flow, Proc. of ASME 2000 Fluids Engineering Division Summer Meeting, Boston, Massachusetts, June 11-15 Toumi I at al (April 2-6, 2000) Development of a multi-dimensional upwind solver for twophase water/steam flows, Proc. of ICONE 8, 8th Int. Conf. On Nuclear Engineering, Baltimore, MD USA Tryggavson G at al. (2001) A front tracking method for the computations of multiphase flows, Journal of Comp. Physics vol 169 pp 708-759
References
635
Verschueren M (1999) A difuse-interface model for structure development in flow, PhD Thesis, Technische Universitteit Eindhoven van Wijngaarden L (1976) Hydrodynamic interaction between gas bubbles in liquid, J. Fluid Mech., vol 77 pp 27-44 Yabe T and Takei E (1988) A New Higher-Order Godunov Method for General Hyperbolic Equations. J.Phys. Soc. Japan, vol 57 p 2598 Xiao F and Yabe T (2001) Completely conservative and oscillation-less semiLagrangian schemes for advection transportation, J. Comput. Phys., vol 170 p 498 Xiao F, Yabe T, Peng X and Kobayashi H, (2002) Conservation and oscillation-less transport schemes based an rational functions, J. Geophys. Res., vol 107, p 4609 Xiao F (2003) Profile-modifiable conservative transport schemes and a simple multi integrated moment formulation for hydrodynamics, in Computational Fluid Dynamics 2002, Amfield S, Morgan P and Srinivas K, eds, Springer, p 106 Xiao F and Ikebata A (2003) An efficient method for capturing free boundary in multifluid simulations, Int. J. Numer. Method in Fluid. vol 4 2 p p 187-210 Yabe T and Wang PY (1991) Unified Numerical Procedure for Compressible and Incompressible Fluid, J. Phys. Soc. Japan vol 60 pp 2105-2108 Yabe T and Aoki A (1991) A Universal Solver for Hyperbolic-Equations by Cubic Polynomial Interpolation. I. One-Dimensional Solver, Comput. Phys. Commun., vol 66 p 219 Yabe T, Ishikawa T, Wang PY, Aoki T, Kadota Y and Ikeda F (1991) A universal solver for hyperbolic-equations by cubic-polynomial interpolation. II. 2-dimensional and 3-dimensional solvers. Comput. Phys. Commun., vol 66 p 233 Yabe T, Xiao F and Utsumi T (2001) Constrained Interpolation Profile Method for Multiphase Analysis, J. Comput. Phys., vol 169 pp 556-593 Yabe T, Tanaka R, Nakamura T and Xiao F (2001) Exactly conservative semiLagrangian scheme (CIP-CSL) in one dimension. Mon. Wea. Rev., vol. 129 p 332 Yabe T, Xiao F and Utsumi T (2001) The constrained interpolation profile method for multiphase analysis, J. Comput. Phys., vol 169 p 556
Crrgpfkz"3"Dtkgh"kpvtqfwevkqp"vq"xgevqt"cpcn{uku"
Before starting to study the theory of multi-phase flows it is advisable to refresh your knowledge on vector analysis. My favorite choice is the book “Calculus and Analytic Geometry” by Thomas et al. (1998). Of course you may use any text book to this topic also. Here only a brief summary is given in order to assist in the understanding of the next chapters of this book. Right handed Cartesian coordinate system: To locate points in space, we use three mutually perpendicular coordinate axes ( x, y , z ) as proposed by Descartes. When you hold your right hand so that the fingers curl from the positive xaxis toward the positive y-axis and your thumb points along the positive z-axis the coordinate systems is right handed. We use right handed coordinate systems. Vector (The Gibbs - Heaviside concept from 1870): A vector in a space is a directed line segment. Two vectors are equal or the same if they have the same length and direction. The vector between two points: The vector from point P1 ( x1 , y1 , z1 ) to point P2 ( x2 , y2 , z2 ) is →
P1P2 = ( x2 − x1 ) i + ( y2 − y1 ) j + ( z2 − z1 ) k .
(1)
Midpoints: The position vector of the midpoint M of the line segment joining points P1 ( x1 , y1 , z1 ) and P2 ( x1 , y1 , z1 ) is rM =
x2 + x1 y + y1 z +z i+ 2 j + 2 1 k. 2 2 2
(2)
Centroid of a triangle: The position vector of the center of mass of a triangle defined by the points P0 ( x0 , y0 , z0 ) , P1 ( x1 , y1 , z1 ) and P2 ( x2 , y2 , z2 ) is rcm =
x2 + x1 + x0 y + y1 + y0 z +z +z i+ 2 j + 2 1 0 k. 3 3 3
(3)
Centroid of the union of non-overlapping plane regions: The Alexandrian Greek Pappus knew in the third century that a centroid of the union of two non-overlapping plane regions lies on the line segment joining their individual
638
Appendix 1 Brief introduction to vector analysis
centroids. More specifically, suppose m1 and m2 are the masses of thin plates P1 and P2 that cover non overlapping regions in the xy plane. Let c1 and c2 be the vectors from the origin to the respective centers of mass of P1 and P2. Then the center of mass of the union P1 ∪ P2 of the two plates is determined by the vector rcm =
m1c1 + m2c 2 . m1 + m2
(4)
This equation is known as the Pappus’s formula. For more than two nonoverlapping plates, as long a their number is finite, the formula generalizes to
∑m c = ∑m
m m
rcm
m
.
(5)
m
m
This formula is especially useful for finding the centroid of a plate of irregular shape that is made up of pieces of constant density whose centroids we know from geometry. We find the centroid of each piece and apply the above equation to find the centroid of the plane. Magnitude: The magnitude (length) of the vector A = a1i + a2 j + a3k is A = a1i + a2 j + a3k = a12 + a22 + a32 .
(6)
Vectors with magnitude equal to one are called unit vectors. Unit vectors are built from direction cosines. Vectors with magnitude zero are called zero vectors and are denoted with 0. Lines and line segments in space: The vector equation for the line through P0 ( x0 , y0 , z0 ) parallel to a vector v = Ai + Bj + Ck is →
P0 P = tv ,
−∞ < t < ∞
(7)
or
( x − x0 ) i + ( y − y0 ) j + ( z − z0 ) k = t ( Ai + Bj + Ck ) ,
−∞ < t < ∞ ,
(8)
where all points P lie on the line and t is a parameter. For line segments t is bounded, e.g. a < t < b . Standard parametrization of the line through P0 ( x0 , y0 , z0 ) parallel to v = Ai + Bj + Ck : The standard parametrization of the line through P0 ( x0 , y0 , z0 )
parallel to the vector v = Ai + Bj + Ck is x = x0 + tA,
y = y0 + tB ,
z = z0 + tC ,
−∞ < t < ∞ .
(9)
Appendix 1 Brief introduction to vector analysis
In other words the points
( x, y , z ) = ( x0 + tA, y0 + tB, z0 + tC )
639
in the interval
−∞ < t < ∞ make up the line.
Position vector defining a line: The vector r ( x, y , z ) = r ( t ) = ( x0 + tA ) i +
( y0 + tB ) j + ( z0 + tC ) k ,
−∞ < t < ∞
(10) from the origin to the point P ( x, y , z ) is the position vector. The position vector defining a line crossing the points P1 ( x1 , y1 , z1 ) and P2 ( x1 , y1 , z1 ) is r ( x, y , z ) = r ( t ) = ⎡⎣ x1 + t ( x2 − x1 ) ⎤⎦ i + ⎡⎣ y1 + t ( y2 − y1 ) ⎤⎦ j + ⎡⎣ z1 + t ( z2 − z1 ) ⎤⎦ k , 0 < t < 1. (11)
Derivative of a position vector defining a line: At any point t the derivative of the position vector is r ( t + Δt ) − r ( t ) dr = lim = = Ai + Bj + Ck , Δ t → 0 dt Δt
(12)
which is parallel to the line. Position vector defining a curve: The vector r ( x, y , z ) = r ( t ) = f ( t ) i + g ( t ) j + h ( t ) k , −∞ < t < ∞
(13)
from the origin to the point P ( x, y , z ) that belongs to the curve is the position vector of the curve. Derivative of a position vector defining a curve: At any point t the derivative of the position vector is the vector r ( t + Δt ) − r ( t ) ∂f dr ∂g ∂h = lim = = i+ j+ k, dt Δt →0 Δt ∂t ∂t ∂t
(14)
- see Fig. A1.1.
The curve traced by r is smooth if dr/dt is continuous and never 0, i.e. if f, g, and h have continuous first derivatives that are not simultaneously 0. The vector dr/dt, when different from 0, is also a vector tangent to the curve. The tangent line through the point ⎡⎣ f ( t0 ) , g ( t0 ) , h ( t0 ) ⎤⎦ is defined to be a line through the point
640
Appendix 1 Brief introduction to vector analysis
dr
r(t )
r (t + dt )
Fig. A1.1 Infinitesimal change of the position vector
parallel to dr/dt at t = t0 . The magnitude of the derivative of the position vector defining the curve is 2
2
2
2
2
2
dr ⎛ df ⎞ ⎛ dg ⎞ ⎛ dh ⎞ ⎛ dx ⎞ ⎛ dy ⎞ ⎛ dz ⎞ = ⎜ ⎟ +⎜ ⎟ +⎜ ⎟ = ⎜ ⎟ +⎜ ⎟ +⎜ ⎟ . dt ⎝ dt ⎠ ⎝ dt ⎠ ⎝ dt ⎠ ⎝ dt ⎠ ⎝ dt ⎠ ⎝ dt ⎠
(15)
For the case where the position vector defines a line between two points P1 ( x1 , y1 , z1 ) and P2 ( x1 , y1 , z1 ) the magnitude of the derivative of the position vector is dr = dt
( x2 − x1 ) + ( y2 − y1 ) + ( z2 − z1 ) 2
2
2
(16)
the distance between the two points. Arc length: The length of a smooth curve r ( t ) = f ( t ) i + g ( t ) j + h ( t ) k , a < t < b , that is traced exactly once as t increases from t = a to t = b is b
2
2
2
2
b
2
2
b
dr ⎛ df ⎞ ⎛ dg ⎞ ⎛ dh ⎞ ⎛ dx ⎞ ⎛ dy ⎞ ⎛ dz ⎞ L = ∫ ⎜ ⎟ + ⎜ ⎟ + ⎜ ⎟ dt = ∫ ⎜ ⎟ + ⎜ ⎟ + ⎜ ⎟ dt = ∫ dt . dt ⎝ dt ⎠ ⎝ dt ⎠ ⎝ dt ⎠ ⎝ dt ⎠ ⎝ dt ⎠ ⎝ dt ⎠ a a a (17)
For the case where the position vector defines a line between two points P1 ( x1 , y1 , z1 ) and P2 ( x2 , y2 , z2 ) the length of the line segment between the two points is 1
L=∫
( x2 − x1 ) + ( y2 − y1 ) + ( z2 − z1 ) 2
2
2
dt =
( x2 − x1 ) + ( y2 − y1 ) + ( z2 − z1 ) 2
2
0
(18)
2
.
Appendix 1 Brief introduction to vector analysis
641
Arc length parameter base point P0 ( x0 , y0 , z0 ) : If we choose P0 ( t0 ) on a smooth curve C parametrized by t, each value of t determines a point P ⎡⎣ x ( t ) , y ( t ) , z ( t )⎤⎦ on C and a “directed distance” t
s (t ) = ∫ t0
dr dτ , dt
(19)
measured along C from the base point. If t > t0 , s ( t ) is the distance from P0 ( t0 ) to P ( t ) . If t < t0 , s ( t ) is the negative of the distance. Each value of s determines a point on C and parametrizes C with respect to s. We call s an arc length parameter for the curve. The parameter’s value increases in direction of increasing t. For a smooth curve the derivatives beneath the radical are continuous and the fundamental theorem of calculus tells us that s is a differentiable function of t with derivative ds dr = . dt dt
(20)
Note that, while the base point P0 ( t0 ) plays a role in defining s, it plays no role in defining ds/dt. Note also that ds / dt > 0 since the curve is smooth and s increases with increasing t. Time derivatives, velocity, speed, acceleration, direction of motion If r ( x, y , z ) = r (τ ) = f (τ ) i + g (τ ) j + h (τ ) k
(21)
is a position vector of a particle moving along a smooth curve in space, and the components are smooth functions of time τ , then v (t ) =
dr dt
(22)
is the particle’s velocity vector, tangent to a curve. At any time τ , the direction of v is the direction of motion. The magnitude of v is the particle’s speed. Speed = v (τ ) =
dr . dτ
(23)
The derivative a=
dv d 2 r = , dτ dτ 2
when it exists, is the particle’s acceleration vector.
(24)
642
Appendix 1 Brief introduction to vector analysis
Differentiation rules for time derivatives: Because the derivatives of vector functions may be computed component by component, the rules for differentiating vector functions have the same form as the rules for differentiating scalar functions. Constant Function Rule: Scalar Multiple Rules: d du df ( fu ) = f + u dτ dτ dτ
dC = 0 (any constant vector C). dτ
d du ( cu ) = c dτ dτ
(25)
(any number c)
(26)
(any differentiable scalar function).
(27)
If u and v are differentiable vector functions of t, then Sum Rule:
d d u dv (u + v) = + . dτ dτ dτ
Difference Rule:
(28)
d du dv (u − v ) = − . dτ dτ dτ
(29)
d du dv (u ⋅ v ) = ⋅ . dτ dτ dτ
(30)
Dot Product Rule:
Cross Product Rule:
d du dv (u × v) = × v + u × dτ dτ dτ
(in that order).
Chain Rule (Short Form): If r is a differentiable function of entiable function of s, then
τ
and
dr dr dτ = . ds dτ ds
(31)
τ
is a differ-
(32)
Unit tangent vector T: The unit tangent vector of a differentiable curve r (t ) is T=
dr dr / dt dr / dt = = , ds ds / dt dr / dt
(33)
T=
dr ⎛ ∂x ∂y ∂z ⎞ =⎜ i+ j+ k ⎟. ds ⎝ ∂s ∂s ∂s ⎠
(34)
or
For the case where the position vector defines a line between two points P1 ( x1 , y1 , z1 ) and P2 ( x2 , y2 , z2 ) the unit tangent vector is T=
( x − x ) i + ( y2 − y1 ) j + ( z2 − z1 ) k . dr dr / dt dr / dt = = = 2 1 2 2 2 ds ds / dt dr / dt ( x2 − x1 ) + ( y2 − y1 ) + ( z2 − z1 )
(35)
Appendix 1 Brief introduction to vector analysis
643
The unit tangent vector plays an important role in the theory of curvilinear coordinate transformation. Curvature of space curve: As a particle moves along a smooth curve in the space, T = dr / ds turns as the curve bends. Since T is a unit vector, its length remains constant and only its direction changes as the particle moves along the curve. The rate at which T turns per unit of the length along the curve is called curvature. The traditional symbol for the curvature is the Greek letter kappa, κ . If T is the tangent vector of a smooth curve, the curvature function of the curve is
κ=
dT . ds
(36)
The principal unit normal vector for space curve: Since T has a constant length, the vector dT/ds is orthogonal to T. Therefore, if we divide dT/ds by its magnitude κ , we obtain a unit vector orthogonal to T. At a point where κ ≠ 0 , the principal unit normal vector for a curve in the space is N=
dT / ds 1 dT = . dT / ds κ ds
(37)
The vector dT/ds points in the direction in which T turns as the curve bends. Therefore, if we face the direction of increasing length, the vector dT/ds points toward the right if T turns clockwise and towards the left if T turns counterclockwise. Because the arc length parameter for a smooth curve is defined with ds/dt positive, ds/dt = ds / dt , the chain rule gives N=
dT / ds ( dT / dt )( dt / ds ) dT / dt = = . dT / ds dT / dt dt / ds dT / dt
(38)
Scalar (dot) product: The scalar product (dot product) A ⋅ B (“A dot B”) of vectors A = a1i + a2 j + a3k ,
(39)
and B = b1i + b2 j + b3k
(40)
is the number A ⋅ B = A B cos θ
(41)
where θ is the angle between A and B. In words, it is the length of A times the length B times the cosine of the angle between A and B. The law of cosines for the 2
2
2
triangle whose sides represent A, B and C is C = A + B − 2 A B cos θ is used to obtain
644
Appendix 1 Brief introduction to vector analysis
A ⋅ B = a1b1 + a2b2 + a3b3 .
(42)
To find the scalar product of two vectors we multiply their corresponding components and add the results. A ⋅ A = a12 + a22 + a32 = A
2
(43)
is called the Euclidian norm. Splitting a vector into components parallel and orthogonal to another vector: The vector B can be split into a component which is parallel to A ⎛ B⋅A ⎞ projA B = ⎜ ⎟A ⎝ A⋅A ⎠
(44)
and a component orthogonal to A, B − projA B , ⎡ ⎛ B⋅A ⎞ ⎛ B⋅A ⎞ ⎤ B=⎜ ⎟ A + ⎢B − ⎜ ⎟ A⎥ . ⎝ A⋅A ⎠ ⎝ A⋅A ⎠ ⎦ ⎣
(45)
Equations for a plane in space: Suppose a plane M passes through a point P0 ( x0 , y0 , z0 ) and is normal (perpendicular) to the non-zero vector n = Ai + Bj + Ck .
(46) →
Then M is the set of all points P ( x, y , z ) for which P0 P is orthogonal to n. That is, P lies on M if and only if →
n ⋅ P0 P = 0 .
(47)
This equation is equivalent to
( Ai + Bj + Ck ) ⋅ ⎡⎣( x − x0 ) i + ( y − y0 ) j + ( z − z0 ) k ⎤⎦ = 0
(48)
A ( x − x 0 ) + B ( y − y 0 ) + C ( z − z0 ) = 0 .
(49)
or
Note that the angle between two intersecting planes is defined to be the (acute) angle determined by their normal vectors. A plane determined by three points: Consider a plane determined by the three points P0 ( x0 , y0 , z0 ) , P1 ( x1 , y1 , z1 ) and P2 ( x2 , y2 , z2 ) . The vector connecting point P0 with point P1 is then →
P0 P1 = ( x1 − x0 ) i + ( y1 − y0 ) j + ( z1 − z0 ) k .
(50)
Appendix 1 Brief introduction to vector analysis
645
The vector connecting point P0 with point P2 is then →
P0 P2 = ( x2 − x0 ) i + ( y2 − y0 ) j + ( z2 − z0 ) k .
(51)
The vector normal to the plane is defined by →
i
→
j
k
P0 P1 × P0 P2 = x1 − x0
y1 − y0
z1 − z0 .
x 2 − x0
y 2 − y0
z 2 − z0
(52)
The plane is then defined by i
j
x1 − x0
y1 − y0
x2 − x0
y 2 − y0
k z1 − z0 ⋅ ⎡⎣( x − x0 ) i + ( y − y0 ) j + ( z − z0 ) k ⎤⎦ = 0 , (53) z 2 − z0
or x0 − x
y0 − y
z0 − z
x1 − x
y1 − y
z1 − z = 0 .
x2 − x
y2 − y
z2 − z
(54)
Laws of the dot product: The dot product is commutative A ⋅ B = B ⋅ A.
(55)
If c is any number (or scalar), then
( cA ) ⋅ B = A ⋅ ( cB ) = c ( A ⋅ B ) .
(56)
If C = c1i + c2 j + c3k is any third vector, then A ⋅ ( B + C) = A ⋅ B + A ⋅ C ,
(57)
that is the dot product obeys the distributive low. Combined with the commutative law it is also evident that
( A + B ) ⋅ C = A ⋅ C + B ⋅ C.
(58)
The last two equations permit us to multiply sums of vectors by the familiar laws of algebra. For example
( A + B) ⋅ (C + D) = A ⋅ C + A ⋅ D + B ⋅ C + B ⋅ D.
(59)
Angle between two non-zero vectors: The angle between two non-zero vectors A and B is
646
Appendix 1 Brief introduction to vector analysis
⎛ A⋅B ⎞ . ⎜ A B ⎟⎟ ⎝ ⎠
θ = arccos ⎜
(60)
Perpendicular (orthogonal) vectors: Non-zero vectors A and B are perpendicular (orthogonal) if and only if A ⋅ B = 0.
(61)
The gradient vector (gradient): The gradient vector (gradient) of the differentiable function f ( x, y , z ) at point P0 ( x0 , y0 , z0 ) is the vector ∇f =
∂f ∂f ∂f i+ j+ k ∂x ∂y ∂z
(62)
obtained by evaluating the partial derivatives of f at P0 ( x0 , y0 , z0 ) . The notation ∇f is read “grad f ” as well as “gradient of f ” and “del f ”. The symbol ∇ by itself is read “del”. Another notation for the gradient is grad f, read the way it is written.
Equations for tangent planes and normal lines: If r ( x, y , z ) = r ( t ) = g ( t ) i + h ( t ) j + k ( t ) k
is a smooth curve on the level surface f ( x, y, z ) = c of a differentiable function f, then f ⎡⎣ g ( t ) , h ( t ) , k ( t )⎤⎦ = c . Differentiating both sides of this equation with respect to t leads to d dc f ⎡ g ( t ) , h ( t ) , k ( t )⎤⎦ = dt ⎣ dt
(63)
df ∂f dg ∂f dh ∂f dk ⎛ ∂f ∂f ∂f ⎞ ⎛ ∂g ∂h ∂k ⎞ = + + =⎜ i+ j+ k⎟⋅⎜ i+ j+ k dt ∂x dt ∂y dt ∂z dt ⎝ ∂x ∂y ∂z ⎠ ⎝ ∂t ∂t ∂t ⎟⎠ dr = ∇f ⋅ = 0. (64) dt
At every point along the curve, ∇f is orthogonal to the curve’s velocity vector dr , and therefore orthogonal to the surface f ( x, y, z ) = c . The unit normal vecdt tor to this surface is n=
∇f . ∇f
If the curve passes through the point P0 ( x0 , y0 , z0 ) :
(65)
Appendix 1 Brief introduction to vector analysis
647
The tangent plane at the point P0 on the level surface f ( x, y, z ) = c is the plane through P0 normal to ∇f
P0
,
⎛ ∂f ∂f ∂f ⎞ j + k ⎟ ⋅ ⎡⎣( x − x0 ) i + ( y − y0 ) j + ( z − z0 ) k ⎤⎦ = 0 ; ⎜ i+ ∂ x ∂ y ∂z ⎠ ⎝
(66)
The normal line of the surface at P0 is the line through P0 parallel to ∇f
P0
.
Gradient of a plane defined by three points: Consider a plane determined by the three points P0 ( x0 , y0 , z0 ) , P1 ( x1 , y1 , z1 ) and P2 ( x2 , y2 , z2 ) x0 − x
y0 − y y1 − y
z1 − z
x2 − x
y2 − y
z2 − z
f ( x, y, z ) = x1 − x
z0 − z
= ( x0 − x )( y1 − y )( z2 − z ) + ( x2 − x )( y0 − y )( z1 − z ) + ( x1 − x )( y 2 − y )( z0 − z ) − ( x2 − x )( y1 − y )( z0 − z ) − ( x1 − x )( y0 − y )( z2 − z ) − ( x0 − x )( y 2 − y )( z1 − z ) = 0
(67) The gradient vector (gradient) of the differentiable function f ( x, y , z ) at point P ( x, y , z ) is the vector ∇f =
∂f ∂f ∂f i+ j+ k, ∂x ∂y ∂z
(68)
with components ∂f = ( y1 − y2 )( z0 − z ) + ( y2 − y0 )( z1 − z ) + ( y0 − y1 )( z2 − z ) , ∂x
(69)
∂f = ( z1 − z2 )( x0 − x ) + ( z2 − z0 )( x1 − x ) + ( z0 − z1 )( x2 − x ) , ∂y
(70)
∂f = ( x1 − x2 )( y0 − y ) + ( x2 − x0 )( y1 − y ) + ( x0 − x1 )( y2 − y ) . ∂z
(71)
→
→
Note that the vectors P0 P1 , P0 P2 , and the vector normal to the plane forms a left oriented vector system. The gradient vector is therefore parallel to the surface normal vector.
648
Appendix 1 Brief introduction to vector analysis
Curvature of a surface: Consider the smooth curves r1 ( x, y, z ) = r1 ( t ) = g1 ( t ) i + h1 ( t ) j + k1 ( t ) k ,
(72)
r2 ( x, y , z ) = r2 ( t ) = g 2 ( t ) i + h2 ( t ) j + k 2 ( t ) k ,
(73)
on the level surface f ( x, y , z ) = c of a differentiable function f. In this case f ⎡⎣ g1 ( t ) , h1 ( t ) , k1 ( t )⎤⎦ = c and f ⎡⎣ g 2 ( t ) , h2 ( t ) , k2 ( t )⎤⎦ = c along the curves. The curves crosses the point P0 ( x0 , y0 , z0 ) . At this point they are orthogonal. The tan-
gents at P0 of these curves are T1 and T2 . s1 and s2 are the arc distances counted from P0 . As a particle moves along the first curve in the space, T1 = dr1 / ds1 turns as the curve bends. Since T1 is a unit vector, its length remains constant and only its direction changes as the particle moves along the curve. The rate at which T1 turns per unit of the length along the curve is called the curvature κ 1 . As T1 is the tangent vector of a smooth curve, the curvature function of the curve is
κ1 =
dT1 = ( T1 ⋅ ∇ ) T1 . ds1
(74)
Similarly, the curvature function of the second curve is
κ2 =
dT2 = ( T2 ⋅ ∇ ) T2 . ds2
(75)
Both curvature functions are called principle curvatures. Their reciprocals are called principal radii. The curvature function of the level surface f ( x, y, z ) = c is defined as
κ = κ1 + κ 2 .
(76)
The gradient along a direction normal and tangential to a surface: Consider the level surface f ( x, y, z ) = c of a differentiable function f with a normal unit vector n=
∇f . ∇f
(77)
The gradient of any differentiable scalar function in space can then be split into a component parallel to the normal vector and a component tangential to the surface: ⎡ ⎛ ∇⋅n ⎞ ⎛ ∇⋅n ⎞ ⎤ ∇=⎜ n + ⎢∇ − ⎜ ⎟ ⎟ n⎥ = ∇n + ∇t , ⎝ n⋅n ⎠ ⎝ n⋅n ⎠ ⎦ ⎣
(78)
Appendix 1 Brief introduction to vector analysis
649
where ∇n = ( ∇ ⋅ n ) n
(79)
and ∇t = ∇ − ∇ n .
(80)
Curvature of a surface defined by the gradient of its unit normal vector: Consider a smooth surface in space defined by its normal vector n ( x, y , z ) = g ( x , y , z ) i + h ( x, y , z ) j + k ( x , y , z ) k.
(81)
In this case there is a convenient method reported by Brackbill et al. (1992) for computation of the curvature of this surface
κ = − ( ∇ ⋅ n ).
(82)
Let us estimate as an example the curvature of a liquid layer in stratified flow between two horizontal planes in the plane y = const. The interface is described by the curve z = δ 2 ( x ) where δ 2 ( x ) is the film thickness. Thus the level surface function is f ( x, y, z ) ≡ δ 2 ( x ) − z = 0 . The gradient of the level surface function is ∇f ( x, y = const , z ) =
∂f ( x, z ) ∂x
i+
∂f ( x, z ) ∂z
k=
∂δ 2 ( x ) ∂x
i−k ,
(83)
and the magnitude of the gradient 1/ 2
⎧ ⎡ ∂δ ( x ) ⎤ 2 ⎫ ⎪ ⎪ ∇f = ⎨1 + ⎢ 2 ⎥ ⎬ ∂x ⎦ ⎪ ⎩⎪ ⎣ ⎭
.
(84)
The normal vector is ⎛ ⎜ ∂δ 2 ( x ) ⎜ ∇f ∂x n2 = − = −⎜ i− 1/ 2 ⎜⎧ 2 ∇f ⎫ ⎡ ⎤ ∂ δ x ( ) ⎪ ⎜ ⎪1 + 2 ⎜⎜ ⎨ ⎢ ∂x ⎥ ⎬ ⎦ ⎭⎪ ⎝ ⎩⎪ ⎣
1 1/ 2
⎧⎪ ⎡ ∂δ ( x ) ⎤ 2 ⎫⎪ 2 ⎨1 + ⎢ ⎥ ⎬ ∂ x ⎦ ⎪ ⎩⎪ ⎣ ⎭
⎞ ⎟ ⎟ k ⎟. ⎟ ⎟ ⎟⎟ ⎠
(85)
Note that for decreasing film thickness with x the normal vector points outside the liquid. This is the result of selecting f ( x, y, z ) ≡ δ 2 ( x ) − z = 0 instead of f ( x, y, z ) ≡ z − δ 2 ( x ) = 0 which is also possible. The curvature is then
650
Appendix 1 Brief introduction to vector analysis
⎛ ⎜ ∂δ 2 ( x ) ⎜ ⎡ ∇f ⎤ ∂x ⎜ κ l = − ( ∇ ⋅ nl ) = ∇ ⋅ ⎢ i− ⎥ = ∇⋅⎜ 1/ 2 ∇ f ⎧⎪ ⎡ ∂δ ( x ) ⎤ 2 ⎫⎪ ⎣⎢ ⎦⎥ ⎜ 1+ 2 ⎜⎜ ⎨ ⎢ ∂x ⎥ ⎬ ⎦ ⎪⎭ ⎝ ⎪⎩ ⎣
1 1/ 2
⎧⎪ ⎡ ∂δ ( x ) ⎤ 2 ⎫⎪ 2 ⎨1 + ⎢ ⎥ ⎬ ∂ x ⎦ ⎪ ⎪⎩ ⎣ ⎭
∂δ 2 ( x ) ∂ ∂ 1 ∂x = − . 1/ 2 1/ 2 2 2 ∂x ⎧ ∂z ⎧ ⎪ ⎡ ∂α 2 ( x ) ⎤ ⎫⎪ ⎪ ⎡ ∂δ 2 ( x ) ⎤ ⎫⎪ ⎨1 + ⎢ ⎨1 + ⎢ ⎥ ⎬ ⎥ ⎬ ∂x ⎦ ⎪ ⎪⎩ ⎣ ∂x ⎦ ⎭⎪ ⎩⎪ ⎣ ⎭
⎞ ⎟ ⎟ k⎟ ⎟ ⎟ ⎟⎟ ⎠
(86)
Having in mind that ∂ 1 =0 1/ 2 2 ∂z ⎧ ⎫ ⎡ ⎤ ∂ δ x ( ) ⎪ ⎪ 2 ⎨1 + ⎢ ⎥ ⎬ ∂ x ⎦ ⎪⎭ ⎪⎩ ⎣
(87)
one finally obtain the well-known expression ∂ 2δ 2 ( x ) κ2 = ∂x 2
⎧⎪ ⎡ ∂δ ( x ) ⎤ 2 ⎫⎪ 2 ⎨1 + ⎢ ⎥ ⎬ ∂ x ⎦ ⎪ ⎪⎩ ⎣ ⎭
3/ 2
.
(88)
Speed of displacement of geometrical surface (Truesdell and Toupin 1960): Consider a set of geometrical surfaces defined by the equation r = r ( t1 , t2 ,τ ) ,
where t1 and t2 are the coordinates of a point on this surface and
(89)
τ
is the time.
The velocity of the surface point ( t1 , t2 ) is defined by ⎛ ∂r ⎞ Vσ = ⎜ ⎟ . ⎝ ∂τ ⎠t1 ,t2
(90)
Expressing the surface equation with f ( x , y , z ,τ ) = 0
(91)
we have ∂f + Vσ ⋅ ∇f = 0. ∂τ
(92)
Appendix 1 Brief introduction to vector analysis
651
Knowing that the outwards directed unit surface vector is n=
∇f ∇f
(93)
the surface velocity can be expressed as Vσ ⋅ n = −
∂f / ∂τ . ∇f
(94)
Potential function: If the vector F is defined on D and F = ∇f
(95)
for some scalar function on D then f is called a potential function for F. The important property of the potential functions is derived from the following consideration. Suppose A and B are two points in D and that the curve r ( x, y , z ) = r ( t ) = f ( t ) i + g ( t ) j + h ( t ) k ,
a ≤ t ≤ b,
(96)
is smooth on D joining A and B. Along the curve, f is a differentiable function of t and df ∂f dx ∂f dy ∂f dz ⎛ ∂f ∂f ∂f ⎞ ⎛ ∂x ∂y ∂z ⎞ = + + =⎜ i+ j+ k⎟⋅ i+ j+ k⎟ dt ∂x dt ∂y dt ∂z dt ⎝ ∂x ∂y ∂z ⎠ ⎜⎝ ∂t ∂t ∂t ⎠ dr dr = ∇f ⋅ = F⋅ . (97) dt dt
Therefore, b
∫ F ⋅ dr = ∫ F ⋅
C
a
b
dr df dt = ∫ dt = f ( A) − f ( B ) , dt dt a
(98)
depends only of the values of f at A and B and not on the path between. In other words if F = M ( x, y, z ) i + N ( x, y, z ) j + P ( x, y, z ) k
(99)
is a field whose component functions have continuous first partial derivatives and there is a solution of the equation M ( x , y , z ) i + N ( x, y , z ) j + P ( x, y , z ) k =
∂f ∂f ∂f i+ j+ k, ∂x ∂y ∂z
(100)
which means solution of the system ∂f = M ( x, y , z ) , ∂x
(101)
652
Appendix 1 Brief introduction to vector analysis
∂f = N ( x, y , z ) , ∂y
(102)
∂f = P ( x, y , z ) , ∂z
(103)
the vector F possesses a potential function f and is called conservative. Differential form, exact differential form: The form M ( x, y, z ) dx + N ( x, y, z ) dy + P ( x, y, z ) dz
(104)
is called the differential form. A differential form is exact on a domain D in space if M ( x, y, z ) dx + N ( x, y, z ) dy + P ( x, y, z ) dz =
∂f ∂f ∂f dx + dy + dz ∂x ∂y ∂z (105)
for some scalar function f through D. In this case f is a smooth function of x, y, and z. Obviously the test for the differential form being exact is the same as the test for F’s being conservative that is the test for the existence of f. Another way for testing whether the differential form is exact is Euler’s relations saying that if the differential form is exact the mutual cross derivatives are equal, that is ∂ ⎛ ∂f ⎞ ∂ ⎛ ∂f ⎞ ⎜ ⎟= ⎜ ⎟, ∂y ⎝ ∂x ⎠ ∂x ⎝ ∂y ⎠
∂ ⎛ ∂ f ⎞ ∂ ⎛ ∂f ⎞ ∂ ⎛ ∂f ⎞ ∂ ⎛ ∂f ⎞ = ⎜ ⎟ , and ⎜ ⎟ ⎜ ⎟ = ⎜ ⎟. ∂z ⎝ ∂x ⎠ ∂x ⎝ ∂z ⎠ ∂y ⎝ ∂z ⎠ ∂z ⎝ ∂y ⎠ (106)
Directional derivatives: Suppose that the function f ( x, y , z ) is defined through the region R in the xyz space, that P0 ( x0 , y0 , z0 ) is a point in R, and that u = u1i + u2 j + u3k is a unit vector. Then the equations x = x0 + su1 ,
y = y0 + su2 ,
z = z0 + su3 ,
−∞ < s < ∞ ,
(107)
parametrize the line through P0 ( x0 , y0 , z0 ) parallel to u. The parameter s measures the arc length from P0 in the direction u. We find the rate of change of f at P0 in the direction u by calculating df/ds at P0 . The derivative of f at P0 in the direction of the unit vector u = u1i + u2 j + u3k is the number f ( x0 + su1 , y0 + su2 , z0 + su3 ) − f ( x0 , y0 , z0 ) ⎛ df ⎞ ⎜ ds ⎟ = lim s →0 s ⎝ ⎠ u, P0
provided the limit exists. Useful relations for practical calculations are
(108)
Appendix 1 Brief introduction to vector analysis
653
⎡⎛ df ⎞ ⎛ df ⎞ ⎛ df ⎞ ⎛ df ⎞ ⎤ ⎜ ds ⎟ = ⎢⎜ dx ⎟ i + ⎜ dy ⎟ j + ⎜ dz ⎟ k ⎥ ⋅ ( u1i + u2 j + u3k ) = ( ∇f ) P0 ⋅ u ⎝ ⎠ u, P0 ⎢⎣⎝ ⎠ P0 ⎝ ⎠ P0 ⎝ ⎠ P0 ⎥⎦ = ∇f ⋅ u cosθ . (109)
At any given point, f increases most rapidly in the direction of ∇f and decreases most rapidly in the direction of - ∇f . In any direction orthogonal to ∇f , the derivative is zero. Derivatives of a function along the unit tangent vector: Suppose that the function f ( x, y , z ) is defined through the region R in the xyz space, and is differentiable, and that T=
dr ⎛ ∂x ∂y ∂z ⎞ =⎜ i+ j+ k⎟ ds ⎝ ∂s ∂s ∂s ⎠
(110)
is a unit tangent vector to the differentiable curve r ( t ) through the region R. P0 ( x0 , y0 , z0 ) is a point at r ( t ) in R. Then the derivative of f at P0 along the
unit tangent vector is the number ∂f ∂f ∂x ∂f ∂y ∂f ∂z ⎛ ∂f ∂f ∂f ⎞ ⎛ ∂x ∂y ∂z ⎞ = + + =⎜ i+ j+ k⎟⋅⎜ i + j+ k⎟ ∂s ∂x ∂s ∂y ∂s ∂z ∂s ⎝ ∂x ∂y ∂z ⎠ ⎝ ∂ s ∂s ∂s ⎠ ∂r = ( ∇f ) ⋅ = ( ∇f ) ⋅ T . (111) ∂s
The vector (cross) products: If A and B are two non-zero vectors in space which are not parallel, they determine a plane. We select a unit vector n perpendicular to the plane by the right-handed rule. This means we choose n to be unit (normal) vector that points the way your right thumb points when your fingers curl through the angle θ from A to B. We then define the vector product A × B (“A cross B”) to be the vector A × B = ( A B sin θ ) n.
(112)
The vector A × B is orthogonal to both A and B because it is a scalar multiple of n. The vector product of A and B is often called the cross product of A and B because of the cross in the notation A × B . Since the sines of 0 and π are both zero in the definition equation, it makes sense to define the cross product of two parallel non-zero vectors to be 0. If one or both of the vectors A and B are zero, we also define A × B to be zero. This way, the cross product of two vectors A and B is zero if and only if A and B are parallel or one or both of them are zero.
654
Appendix 1 Brief introduction to vector analysis
Note that A × B = − (B × A) .
(113)
The determination formula for B = b1i + b2 j + b3k , then i
j
A × B = a1
a2
a3 .
b1
b2
b3
A × B : If
A = a1i + a2 j + a3k , and
k
(114)
As an example consider the expression appearing in the multi-phase fluid mechanics for computation of the lift forces ( Vl − Vm ) × ( ∇ × Vm ) as a function of the velocity vectors. The result is given in two steps
( ∇ × Vm ) =
i
j
∂ ∂x um
∂ ∂y vm
k ⎛ ∂w ∂ ∂v ⎞ ⎛ ∂w ∂u ⎞ ⎛ ∂v ∂u ⎞ = ⎜ m − m ⎟ i − ⎜ m − m ⎟ j+ ⎜ m − m ⎟ k, ∂z ⎝ ∂y ∂z ⎠ ⎝ ∂x ∂z ⎠ ⎝ ∂x ∂y ⎠ wm
(115)
( Vl − Vm ) × ( ∇ × Vm ) = ⎡ ⎛ ∂v ∂u ⎞ ∂u ⎞ ⎤ ⎛ ∂w = ⎢( vl − vm ) ⎜ m − m ⎟ + ( wl − wm ) ⎜ m − m ⎟ ⎥ i ∂y ⎠ ∂z ⎠ ⎦ ⎝ ∂x ⎝ ∂x ⎣ ⎡ ⎛ ∂v ⎛ ∂w ∂u ⎞ ∂v ⎞ ⎤ − ⎢( ul − um ) ⎜ m − m ⎟ − ( wl − wm ) ⎜ m − m ⎟ ⎥ j ∂y ⎠ ∂z ⎠ ⎦ ⎝ ∂x ⎝ ∂y ⎣ ⎡ ⎛ ∂w ∂u ⎞ ∂v ⎞ ⎤ ⎛ ∂w + ⎢ − ( ul − um ) ⎜ m − m ⎟ − ( vl − vm ) ⎜ m − m ⎟ ⎥ k . ∂z ⎠ ∂z ⎠ ⎦ ⎝ ∂x ⎝ ∂y ⎣
(116)
Parallel vectors: Non-zero vectors A and B are parallel if and only if A × B = 0.
(117)
The associative and distributive lows for the cross product: The scalar distributive law is
( rA ) × ( sB ) = ( rs )( A × B ) ,
(118)
with the special case
( − A ) × B = A × ( −B ) = − ( A × B ) .
(119)
Appendix 1 Brief introduction to vector analysis
655
The vector distributive laws are A × ( B + C ) = A × B + A × C,
(120)
( B + C) × A = B × A + C × A .
(121)
Note the interesting vector identities, Thompson et al. (1985), p. 100 and 101,
( A × B ) ⋅ ( C × D ) = ( A ⋅ C )( B ⋅ D ) − ( A ⋅ D )( B ⋅ C ) ,
(122)
and A × ( B × C) = ( A ⋅ C) B − ( A ⋅ B ) C .
(123)
The area of a parallelogram: Because n is a unit vector, the magnitude of A × B is A × B = A B sin θ n = A B sin θ .
(124)
This is the area of the parallelogram determined by A and B, A being the base of the parallelogram and B sin θ the height. Area of union of two triangle plane regions: Consider four vertices points defined by the position vector ri = xi i + yi j + zi k ,
(125)
where i = 5, 6, 8, 7, see Fig. A1.2.
S 5 = S 5687 S 3 = S 3487
7
S 2 = S1375
8 5
ζ
z
S1 = S 2486
3
4
S 4 = S1265
6
η
y 1
ξ x
2
S 6 = S1243
Fig. A1.2 Tetrahedron defined by joint triangles
656
Appendix 1 Brief introduction to vector analysis
In general the four points must not belong to the same plane. The surface vector defined by these points is computed as follows. First we compute the surface vectors of the triangles 587 and 568 S587 =
1 ( r5 − r7 ) × ( r8 − r7 ), 2
(126)
S568 =
1 ( r6 − r8 ) × ( r5 − r6 ) . 2
(127)
The vector sum of the above two vectors yields an expression for the vector surface S5687 =
1 ( r6 − r5 ) × ( r8 − r5 ), 2
(128)
Kordulla and Vinokur (1983). In fact this is the surface defined by the vector product of the vectors joining the couple of the opposite vertices, respectively. One should note that the first, the second and the resulting vector form a right handed coordinate systems in order to specify the outward direction of the surface if it is part of the control volume surface. Thus, for the hexadron presented in Fig. A1.2 we have S1 = S 2486 =
1 ( r4 − r6 ) × (r8 − r2 ) , 2
(129)
S 2 = S1375 =
1 (r3 − r5 ) × ( r1 − r7 ) , 2
(130)
S3 = S3487 =
1 ( r4 − r7 ) × ( r3 − r8 ), 2
(131)
S 4 = S1265 =
1 (r6 − r1 ) × (r5 − r2 ) , 2
(132)
S5 = S5687 =
1 (r8 − r5 ) × ( r7 − r6 ) , 2
(133)
S6 = S1243 =
1 ( r4 − r1 ) × ( r2 − r3 ) . 2
(134)
The triple scalar or box product: The product ( A × B ) ⋅ C is called the triple scalar product of A, B, and C (in that order). As you can see from the formula
(A × B) ⋅ C =
A × B C cos θ ,
(135)
the absolute value of the product is the volume of the parallelepiped (parallelogram-sided box) determined by A, B, and C. The number A × B is the area of the
Appendix 1 Brief introduction to vector analysis
657
parallelogram. The number C cos θ is the parallelogram’s height. Because of the geometry ( A × B ) ⋅ C is called the box product of A, B, and C. By treating the planes of B and C and of C and A as the base planes of the parallelepiped determined by A, B, and C, we see that
( A × B ) ⋅ C = ( B × C) ⋅ A = (C × A ) ⋅ B.
(136)
Since the dot product is commutative, the above equation gives
( A × B ) ⋅ C = A ⋅ (B × C).
(137)
The triple scalar product can be evaluated as a determinant
( A × B) ⋅ C =
a1 a2
a3
b1
b2
b3 .
c1
c2
c3
(138)
Volume of a tetrahedron: Consider a tetrahedron defined by the four vertices numbered as shown in Fig. A1.3.
5 3
ζ
z
η
y 1
ξ
2
x Fig. A1.3 Definition of the vertices numbering of the tetrahedron
The volume of the tetrahedron is then V =
1 ( r5 − r1 ) ⋅ ⎡⎣( r2 − r1 ) × ( r3 − r1 )⎤⎦ . 6
(139)
Volume of a hexahedron: Consider a hexahedron defined by the eight vertices numbered starting with base in the counterclockwise direction as shown in Fig. A1.4.
658
Appendix 1 Brief introduction to vector analysis
7 8 5
ζ
z
3
4
6
η
y 1
ξ
2
x Fig. A1.4 Definition of the vertices numbering of the hexahedron
An alternative approach for computation of the volume of the hexahedron is to compute the volumes of the tetrahedra making up the hexahedron. Kordulla and Vinokur (1983) used symmetric partitioning of the faces decomposing the hexahedron into four tetrahedra and obtained V12435687 = −
1 1 ( r8 − r1 ) ⋅ ( S1243 + S1265 + S1375 ) = − ( r8 − r1 ) ⋅ ( S2 + S4 + S6 ) . 3 3 (140)
Here the surfaces S1243 , S1265 , and S1375 belongs to the same vertices and r8 − r1 is the diagonal joining these vertices with the opposite ones. Dyadic product of two vectors: Consider the vectors A = a1i + a2 j + a3k , B = b1i + b2 j + b3k . The dyadic product of the two vectors is written as AB without any sign between them, and is defined as the following second order tensor ⎛ a1b1 ⎜ AB = ⎜ a2 b1 ⎜a b ⎝ 3 1
a1b2 a2 b2 a3b2
a1b3 ⎞ ⎟ a2b3 ⎟ . a3b3 ⎟⎠
(141)
Note that the dyadic product of AB is not equivalent to BA. As examples we give ∂ ∂ ∂ some dyadic products between the vectors ∇⋅ = i+ j + k and ∂x ∂y ∂z V = ui + vj + vk , which are used in fluid mechanics for definition of the friction forces in flows:
Appendix 1 Brief introduction to vector analysis
⎛ ∂u ⎜ ⎜ ∂x ⎜ ∂u ∇V = ⎜ ∂y ⎜ ∂ ⎜ u ⎜ ⎝ ∂z
( ∇V )
T
∂v ∂x ∂v ∂y ∂v ∂z
⎛ ∂u ⎜ ⎜ ∂x ⎜ ∂v =⎜ ⎜ ∂x ⎜ ∂w ⎜ ∂x ⎝
∂w ⎞ ⎟ ∂x ⎟ ∂w ⎟ , ∂y ⎟ ⎟ ∂w ⎟ ⎟ ∂z ⎠ ∂u ∂y ∂v ∂y ∂w ∂y
659
(142)
∂u ⎞ ⎟ ∂z ⎟ ∂v ⎟ ⎟. ∂z ⎟ ∂w ⎟ ∂z ⎟⎠
(144)
Another dyadic product of the velocity vector ⎛ uu uv uw ⎞ ⎜ ⎟ VV = ⎜ vu vv vw ⎟ , ⎜ wu wv ww ⎟ ⎝ ⎠
(145)
is used to describe the convective momentum transport. Eigenvalues: The eigenvalues λi of a matrix A are the solution of the characteristic polynomial A − λ I = det ( A − λ I ) = 0 ,
(146)
where I is the identity matrix. Eigenvectors: The right eigenvector of a matrix A corresponding to a an eigenvalue λi of A is a vector K ( ) = ⎡ k1( ) , k2( ) ,..., km( ) ⎤ ⎣ ⎦ i
i
i
i
AK ( ) = λi K ( ) . i
i
T
satisfying (147)
Similarly, a left eigenvector of a matrix A corresponding to a an eigenvalue λi of A is a vector L( ) = ⎡ k1( ) , k2( ) ,..., km( ) ⎤ ⎣ ⎦ i
AL( ) = λi L( ) . i
i
i
i
i
T
satisfying (148)
Diagonalizable matrix: A matrix A is said to be diagonalizable if A can be expressed as A = KHK −1 ,
(149)
660
Appendix 1 Brief introduction to vector analysis
in terms of a diagonal matrix H and a matrix K. The diagonal element of H are the eigenvalues λi of A and the columns of K are the right eigenvectors of A corresponding to the eigenvalues λi , that is ⎡λ1 ⎢0 H=⎢ ⎢. ⎢ ⎣⎢ 0
0
λ2 . 0
0⎤ ⎥ T . 0⎥ , K = ⎡⎣ K (i ) , K (i ) ,..., K (i ) ⎤⎦ ⎥ . . ⎥ . λm ⎦⎥ .
, AK (i ) = λi K (i ) . (150)
Rotation around axes: Given a point ( x0 , y0 , z0 ) . A new position of this point
( x, y , z )
obtained after rotation by angle ϕ with respect to the x, y or z axes re-
spectively is ⎛ x⎞ ⎛ x0 ⎞ ⎜ ⎟ ⎜ ⎟ ⎜ y ⎟ = R i ⎜ y0 ⎟ , ⎜z⎟ ⎜z ⎟ ⎝ ⎠ ⎝ 0⎠
(151)
where 0 ⎛1 ⎜ R1 = ⎜ 0 cos ϕ ⎜ 0 sin ϕ ⎝
0 ⎞ ⎛ cos ϕ ⎟ ⎜ − sin ϕ ⎟ , R 2 = ⎜ 0 ⎜ − sin ϕ cos ϕ ⎟⎠ ⎝
0 sin ϕ ⎞ ⎛ cos ϕ − sin ϕ 0 ⎞ ⎟ ⎜ ⎟ 0 ⎟ , R 3 = ⎜ sin ϕ cos ϕ 0 ⎟ . ⎜ 0 0 cos ϕ ⎟⎠ 0 1 ⎟⎠ ⎝ (152,153,154)
1
Scaling: Given a point ( x0 , y0 , z0 ) . A new position of this point ( x, y, z ) obtained after scaling by ⎛ sx ⎜ S1 = ⎜ 0 ⎜0 ⎝
0 sy 0
0⎞ ⎟ 0 ⎟, s z ⎟⎠
(155)
is ⎛ x⎞ ⎛ x0 ⎞ ⎜ ⎟ ⎜ ⎟ ⎜ y ⎟ = Si ⎜ y0 ⎟ . ⎜z⎟ ⎜z ⎟ ⎝ ⎠ ⎝ 0⎠
(156)
Translation: Graphical systems make use of the general notation of transformation operation defining by multiplying the coordinates with a 4 by 4 matrix
Appendix 1 Brief introduction to vector analysis
⎛x⎞ ⎛ x0 ⎞ ⎜ ⎟ ⎜ ⎟ ⎜ y ⎟ = T ⎜ y0 ⎟ , ⎜z⎟ ⎜ z0 ⎟ ⎜⎜ w ⎟⎟ ⎜⎜ 1 ⎟⎟ ⎝ ⎠ ⎝ ⎠
661
(157)
called translation matrix ⎛1 ⎜ 0 T=⎜ ⎜0 ⎜⎜ 0 ⎝
0 0 Δx ⎞ ⎟ 1 0 Δy ⎟ . 0 1 Δz ⎟ ⎟ 0 0 1 ⎟⎠
(158)
w is used for hardware scaling in graphical representations. Replacing the upper 3 by 3 elements with the matrices defining rotation or scaling results in general notation of transformation. Then multiple transformations on objects are simply sequences of matrices multiplications. Note that the round off errors may damage the objects coordinates after many multiplications. Therefore one should design systems by starting the transformations always from a initial state. References Brackbill JU, Kothe DB, Zemach C (1992) A continuum method for modeling surface tension, Journal of Computational Physics, vol 100 pp 335-354 Kordulla W, Vinokur M (1983) AIAA J., vol 21 pp 917-918 Thomas GB Jr, Finney RL, Weir MD (1998) Calculus and analytic geometry, 9th Edition, Addison-Wesley Publishing Company, Reading, Massachusetts etc. Thompson JF, Warsi ZUA, Mastin CW (1985) Numerical grid generation, North-Holand, New York-Amsterdam-Oxford Truesdell CA, Toupin RA (1960) The classical field theories, Encyclopedia of Physics, S. Fluge Ed., III/1, Principles of Classical Mechanics and Field Theory, Springer, pp 226-858 Vinokur M (1974) Conservation equations of gas dynamics in curvilinear coordinate systems, J. Comput. Phys., vol 14, pp 105-125
Crrgpfkz"4"Dcukeu"qh"vjg"eqqtfkpcvg" vtcpuhqtocvkqp"vjgqt{"
Cartesian coordinate systems are usually the first choice of the scientist and the engineer. Unfortunately in nature and technology the boundaries of a flow may have a very complicated form. Prescribing physically adequate boundary condition at such surfaces is frequently not a solvable problem. The next choice of course is one of the well-known curvilinear orthogonal coordinate systems, polar, or bipolar, or cylindrical. The purpose of this Appendix is to collect from the literature the most important basics on which the general coordinate transformation theory relies. The material is ordered in a generic way. Each statement follows from the already defined statement or statements. For those who would like to perfect his knowledge in this field there are two important books that are my favorite choices are Thomas et al. (1998) - about the calculus and analytic geometry, and Thompson et al. (1985) about numerical grid generation. For a complete collection of basic definitions, rules and useful formula see Miki and Takagi (1984).
ζ
z
η
y
ξ x Fig. A2.1 Cartesian and curvilinear coordinate systems
Consider three non identical and non-parallel curves in the space ξ ,η , ζ having only one common point designated as an origin and presented in Fig. A2.1. The curves are smooth (al least one times differentiable). We call these curves coordinate lines or curvilinear coordinates. The curvilinear coordinate lines of a three-dimensional system can be considered also as space curves formed by the
664
Appendix 2 Basics of the coordinate transformation theory
intersection of surfaces on which one of the coordinates is hold constant. One coordinate varies along a coordinate line, while the other two are constant thereon. Thus, we have the coordinate transformation defined by x = f (ξ ,η , ζ ),
(1)
y = g (ξ ,η , ζ ),
(2)
z = h (ξ ,η , ζ ).
(3)
The surfaces defined by ξ = const , η = const , ζ = const , are called coordinate surfaces. We talk then of physical and transformed (computational) space. In the physical space the coordinates ( x, y, z ) become independent variables. In the transformed space the transformed computational coordinates (ξ ,η , ζ ) become independent variables.
a3
z
a2 y
a1 x Fig. A2.2 The covariant tangent vectors to the coordinate lines form the so called covariant base vectors of the curvilinear coordinate system
The tangent vectors to the coordinate lines form the so called base vectors of the coordinate system. These base vectors are called covariant base vectors, Fig. A2.2. The normal vectors to the coordinate surfaces form the so called contravariant base vectors. The two types of the base vectors are illustrated in Fig. A2.3, showing an element of volume with six sides, each of which lies on some coordinate surface. Another frequently used notation is xi
( i = 1, 2,3) :
x1 , x2 , x3 and ξ i
( i = 1, 2,3) :
ξ , ξ , ξ . In the latter case the superscripts serve only as labels and not as powers. 1
2
3
Appendix 2 Basics of the coordinate transformation theory
665
a3 a2
ζ = const
z
η = const
y
a1
ξ = const x
Fig. A2.3 The contravariant normal vectors to the coordinate surfaces form the so called contravariant base vectors of the curvilinear coordinate system
Metrics and inverse metrics: The differential increments in the transformed coordinate system as a function of the differential increments in the physical coordinate system can be computed as follows ⎛ ∂ξ ⎜ ∂x ⎛ dξ ⎞ ⎜ ⎜ ⎟ ⎜ ∂η ⎜ dη ⎟ = ⎜ ∂x ⎜ dζ ⎟ ⎜ ⎝ ⎠ ⎜ ∂ζ ⎜⎜ ⎝ ∂x
∂ξ ∂y ∂η ∂y ∂ζ ∂y
∂ξ ⎞ ⎟ ∂z ⎟ ∂η ⎟ ⎟ ∂z ⎟ ∂ζ ⎟ ⎟ ∂z ⎟⎠
⎛ dx ⎞ ⎜ ⎟ ⎜ dy ⎟ . ⎜ dz ⎟ ⎝ ⎠
(4)
The elements of the matrix in Eq. (4) are called metrics of the coordinate transformation, Anderson (1995) p.178. An alternative notation of the metrics is aij. The differential increments in the physical coordinate system are a function of the differential increments in the transformed coordinate system ⎛ ∂x ⎜ ∂ξ ⎛ dx ⎞ ⎜ ⎜ ∂y ⎜ ⎟ ⎜ dy ⎟ = ⎜ ∂ξ ⎜ dz ⎟ ⎜ ⎝ ⎠ ⎜ ∂z ⎜⎜ ⎝ ∂ξ
where
∂x ∂η ∂y ∂η ∂z ∂η
∂x ∂ζ ∂y ∂ζ ∂z ∂ζ
⎞ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟⎟ ⎠
⎛ dξ ⎞ ⎛ dx ⎞ ⎛ dξ ⎞ ⎜ ⎟ ⎜ ⎟ ⎜ ⎟ ⎜ dη ⎟ or ⎜ dy ⎟ = J ⎜ dη ⎟ . ⎜ dζ ⎟ ⎜ dz ⎟ ⎜ dζ ⎟ ⎝ ⎠ ⎝ ⎠ ⎝ ⎠
(5)
666
Appendix 2 Basics of the coordinate transformation theory
⎛ ∂x ⎜ ⎜ ∂ξ ⎜ ∂y J (ξ , η , ζ ) = ⎜ ⎜ ∂ξ ⎜ ∂z ⎜⎜ ⎝ ∂ξ
∂x ∂η ∂y ∂η ∂z ∂η
∂x ∂ζ ∂y ∂ζ ∂z ∂ζ
⎞ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟⎟ ⎠
(6)
is the so called Jacobian matrix of the coordinate transformation x = f (ξ ,η , ζ ) , y = g (ξ ,η , ζ ) , z = h (ξ ,η , ζ ) . The elements of the Jacobian matrix are called in-
verse metrics of the coordinate transformation Anderson (1995). An alternative notation of the inverse metrics is aijT . The relation between the so called metrics and the inverse metrics is obvious by noting that the transformation is invertable ⎛ ∂x ⎜ ∂ξ ⎛ dξ ⎞ ⎜ ⎜ ∂y ⎜ ⎟ ⎜ dη ⎟ = ⎜ ∂ξ ⎜ dζ ⎟ ⎜ ⎝ ⎠ ⎜ ∂z ⎜⎜ ∂ξ ⎝
∂x ∂η ∂y ∂η ∂z ∂η
∂x ∂ζ ∂y ∂ζ ∂z ∂ζ
⎞ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟⎟ ⎠
−1
⎛ ∂ξ ⎜ ∂x ⎛ dx ⎞ ⎜ ⎜ ∂ η ⎜ ⎟ ⎜ dy ⎟ = ⎜ ∂x ⎜ dz ⎟ ⎜ ⎝ ⎠ ⎜ ∂ζ ⎜⎜ ∂x ⎝
∂ξ ∂y ∂η ∂y ∂ζ ∂y
∂ξ ⎞ ⎟ ∂z ⎟ ∂η ⎟ ⎟ ∂z ⎟ ∂ζ ⎟ ⎟ ∂z ⎟⎠
⎛ dx ⎞ ⎜ ⎟ ⎜ dy ⎟ , ⎜ dz ⎟ ⎝ ⎠
(7)
∂y ∂x ∂y ⎞ − ⎟ ∂ζ ∂ζ ∂η ⎟ ∂y ∂x ∂y ⎟ − ⎟, ∂ξ ∂ξ ∂ζ ⎟ ∂y ∂x ∂ y ⎟ − ⎟ ∂η ∂η ∂ξ ⎟⎠
(8)
and therefore
J −1
⎛ ∂ξ ⎜ ⎜ ∂x ⎜ ∂η =⎜ ⎜ ∂x ⎜ ∂ζ ⎜⎜ ∂x ⎝
⎛ ∂y ⎜ ⎜ ∂η 1 ⎜ ∂y = ⎜ g ⎜ ∂ζ ⎜ ∂y ⎜⎜ ⎝ ∂ξ
∂ξ ∂y ∂η ∂y ∂ζ ∂y
∂ξ ⎞ ⎛ ∂x ⎟ ⎜ ∂z ⎟ ⎜ ∂ξ ∂η ⎟ ⎜ ∂y ⎟=⎜ ∂z ⎟ ⎜ ∂ξ ∂ζ ⎟ ⎜ ∂z ⎟ ⎜ ∂z ⎟⎠ ⎜⎝ ∂ξ
∂z ∂y ∂z − ∂ζ ∂ζ ∂η ∂z ∂y ∂z − ∂ξ ∂ξ ∂ζ ∂z ∂y ∂ z − ∂η ∂η ∂ξ
where g = J (ξ , η , ζ )
∂x ∂η ∂y ∂η ∂z ∂η
∂x ∂ζ ∂y ∂ζ ∂z ∂ζ
⎞ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟⎟ ⎠
∂x ∂z ∂x ∂z − ∂ζ ∂η ∂η ∂ζ ∂x ∂z ∂x ∂z − ∂ξ ∂ζ ∂ζ ∂ξ ∂x ∂z ∂x ∂z − ∂η ∂ξ ∂ξ ∂η
−1
∂x ∂η ∂x ∂ζ ∂x ∂ξ
Appendix 2 Basics of the coordinate transformation theory
=
∂x ⎛ ∂y ∂z ∂y ∂z ⎞ ∂x ⎛ ∂y ∂z ∂y ∂z ⎞ ∂x ⎛ ∂y ∂z ∂y ∂z − − − ⎜ ⎟− ⎜ ⎟+ ⎜ ∂ξ ⎝ ∂η ∂ζ ∂ζ ∂η ⎠ ∂η ⎝ ∂ξ ∂ζ ∂ζ ∂ξ ⎠ ∂ζ ⎝ ∂ξ ∂η ∂η ∂ξ (9)
667
⎞ ⎟, ⎠
is called the Jacobian determinant or Jacobian of the coordinate transformation. An alternative notation of the metrics is aij. We see that
( )
J = aijT = a ij
−1
,
( )
aij = J −1 = aijT
( )
aij = ⎡ a ij ⎢⎣
−1
T
−1
(10) ,
(11)
( )
⎤ = ⎡ a ij ⎥⎦ ⎢⎣
T
−1
⎤ . ⎥⎦
(12)
Covariant base vectors: The tangent vectors ( a1 , a 2 , a3 ) to the three curvilinear coordinate lines represented by (ξ ,η , ζ ) are called the three covariant base vectors of the curvilinear coordinate system, and are designated with a1 :
∂r ∂x ∂y ∂z = i+ j+ k, ∂ξ ∂ξ ∂ξ ∂ξ
(13)
a2 :
∂r ∂x ∂y ∂z = i+ j+ k, ∂η ∂η ∂η ∂η
(14)
a3 :
∂r ∂x ∂y ∂z = i+ j+ k. ∂ζ ∂ζ ∂ζ ∂ζ
(15)
The components of the three covariant base vectors of the curvilinear coordinate system are the columns of the matrix of the inverse metrics (Jacobian matrix – Eq. 6). They are not unit vectors. The corresponding unit vectors are e1 =
a1 a a , e 2 = 2 , e3 = 3 . a1 a2 a3
(16,17,18)
Contravariant base vectors: The normal vectors to a coordinate surface on which the coordinates ξ , η and ζ are constant, respectively are given by a1 :
∇ξ =
∂ξ ∂ξ ∂ξ i+ j+ k, ∂x ∂y ∂z
(19)
a2 :
∇η =
∂η ∂η ∂η i+ j+ k, ∂x ∂y ∂z
(20)
668
Appendix 2 Basics of the coordinate transformation theory
∇ζ =
a3 :
∂ζ ∂ζ ∂ζ i+ j+ k. ∂x ∂y ∂z
(21)
The components of the three contravariant base vectors of the curvilinear coordinate system are the rows of the matrix of the metrics – Eq. (4). They are not unit vectors. The corresponding unit vectors are
ξ = const ,
e1 =
η = const ,
e2 =
ζ = const ,
e3 =
a1 a1 a2 a2 a3 a3
,
(22)
,
(23)
.
(24)
At any given point on the surface defined with ξ = const , ξ increases most rapidly in the direction of ∇ξ and decreases most rapidly in the direction of - ∇ξ . In any direction orthogonal to ∇ξ , the derivative is zero. The corresponding statements are valid for the remaining surfaces η = const and ζ = const . These normal vectors to the three coordinate surfaces are called the three contravariant base vectors of the curvilinear coordinate system. Cartesian vector in co- and contravariant coordinate systems: Any Cartesian vector V = ui + vj + wk has two sets of distinct components with respect to the frames of the covariant and contravariant base vectors V = V ′1a1 + V ′2 a 2 + V ′3a 3 ,
(25)
V = V1′a1 + V2′a 2 + V3′a3 .
(26)
The components V ′ and Vi′ are called contravariant and covariant components i
of V respectively. The components are easily computed by equalizing the Cartesian components as follows ui + vj + wk = V ′1a1 + V ′2 a 2 + V ′3 a 3
( + (V ′ a
) ( ) k,
)
= V ′1a11 + V ′2 a21 + V ′3 a31 i + V ′1a12 + V ′2 a22 + V ′3 a32 j 1
13
or
+ V ′ a23 + V ′ a33 2
3
(27)
Appendix 2 Basics of the coordinate transformation theory
⎛ a11 ⎜ ⎜ a12 ⎜a ⎝ 13
a31 ⎞ ⎛ V ′1 ⎞ ⎛ u ⎞ ⎟ ⎜ ⎟ ⎟⎜ a32 ⎟ ⎜ V ′2 ⎟ = ⎜ v ⎟ a33 ⎟⎠ ⎜⎝ V ′3 ⎟⎠ ⎜⎝ w ⎟⎠
a21 a22 a23
⎛ V ′1 ⎞ ⎛ u ⎞ ⎜ ⎟ ⎜ ⎟ J ⎜ V ′2 ⎟ = ⎜ v ⎟ ⎜ 3⎟ ⎜ ⎟ ⎜V ′ ⎟ ⎝ w⎠ ⎝ ⎠
or
669
(28)
or after comparing with Eq. (9) we have ⎛ V ′1 ⎞ ⎛ a11 ⎜ 2⎟ ⎜ ⎜ V ′ ⎟ = ⎜ a12 ⎜ V ′3 ⎟ ⎜ a ⎝ ⎠ ⎝ 13
a21 a22 a23
a31 ⎞ ⎟ a32 ⎟ a33 ⎟⎠
−1
⎛u⎞ ⎜ ⎟ −1 ⎜ v ⎟ = J ⋅ V. ⎜ w⎟ ⎝ ⎠
(29)
Therefore V ′1 = a11u + a12 v + a13 w = a1 ⋅ V ,
(30)
V ′2 = a 21u + a 22 v + a 23 w = a 2 ⋅ V ,
(31)
V ′3 = a 31u + a 32 v + a 33 w = a 3 ⋅ V ,
(32)
and Eq. (25) can be rewritten as follows
(
)
(
)
(
)
V = a1 ⋅ V a1 + a 2 ⋅ V a 2 + a 3 ⋅ V a 3 ,
(33)
compare with Eq. (35) in Thompson, Warsi and Mastin (1985) p. 109. Similarly we have 11 ⎛ V1′⎞ ⎛ a ⎜ ⎜ ′⎟ 12 ⎜ V2 ⎟ = ⎜ a ⎜ V ′ ⎟ ⎜ a13 ⎝ 3⎠ ⎝
a 21 a 22 a
23
a 31 ⎞ ⎟ a 32 ⎟ a 33 ⎟⎠
−1
⎛u⎞ ⎜ ⎟ ⎡ ij ⎜ v ⎟ = ⎢⎣ a ⎜ w⎟ ⎝ ⎠
( )
T
⎤ ⎥⎦
−1
( )
= ⎡ a ij ⎢⎣
−1
T
⎤ = aT = J ⋅ V , ij ⎥⎦
(34)
and V = ( a1 ⋅ V ) a1 + ( a 2 ⋅ V ) a 2 + ( a 3 ⋅ V ) a 3,
(35)
compare with Eq. (36) in Thompson, Warsi and Mastin (1985) p. 109. Remember again that the components ai ⋅ V and ai ⋅ V are called contravariant and covariant components of V respectively. Increment of the position vector: The infinitesimal increment of the position vector is dr = a1dξ + a 2 dη + a3 dζ .
(36)
Increment of the arc length: The infinitesimal increment of the arc length along a general space curve can be approximated with the magnitude if the infinitesimal increment of the position vector
670
Appendix 2 Basics of the coordinate transformation theory
ds = dr = dr ⋅ dr
(37)
or ds = dr ⋅ dr = ( a1d ξ + a 2 dη + a3 dζ ) ⋅ ( a1d ξ + a 2 dη + a3 dζ ) 2
⎛ ( a1 ⋅ a1 ) dξ dξ + ( a 2 ⋅ a1 ) dξ dη + ( a3 ⋅ a1 ) dξ dζ ⎜ ⎜ = ⎜⎜ + ( a1 ⋅ a 2 ) dη d ξ + ( a 2 ⋅ a 2 ) dη dη + ( a3 ⋅ a 2 ) dη dζ ⎜ ⎜ ⎜ + ( a1 ⋅ a3 ) d ζ dξ + ( a 2 ⋅ a3 ) d ζ dη + ( a3 ⋅ a3 ) dζ dζ ⎝
⎞ ⎟ ⎟ ⎟. ⎟ ⎟ ⎟ ⎟ ⎠
(38)
The tensor built by the dot products is obviously symmetric, and is called the covariant tensor. Covariant metric tensor: The covariant metric tensor is defined as ⎛ a1 ⋅ a1 a1 ⋅ a 2 ⎜ g ij = ⎜ a 2 ⋅ a1 a 2 ⋅ a 2 ⎜ a ⋅a a ⋅a 3 2 ⎝ 3 1
a1 ⋅ a3 ⎞ ⎟ a 2 ⋅ a3 ⎟ , a3 ⋅ a3 ⎟⎠
(39)
which is a symmetric tensor. Angles between the unit covariant vectors: The angles between the covariant unit vectors can be expressed in terms of the elements of the covariant metric tensor as follows ⎛ g ⎛ a1 ⋅ a 2 ⎞ 12 = arccos ⎜ ⎟ ⎜ a1 a 2 ⎟ ⎜ g g ⎝ ⎠ ⎝ 11 22
⎞ ⎟, ⎟ ⎠
(40)
⎛ g ⎛ a1 ⋅ a3 ⎞ 13 = arccos ⎜ ⎜ a1 a3 ⎟⎟ ⎜ g11 g33 ⎝ ⎠ ⎝
⎞ ⎟, ⎟ ⎠
(41)
⎛ g ⎛ a 2 ⋅ a3 ⎞ 23 = arccos ⎜ ⎜ a 2 a3 ⎟⎟ ⎜ g 22 g33 ⎝ ⎠ ⎝
⎞ ⎟. ⎟ ⎠
(42)
θ12 = arccos ⎜
θ13 = arccos ⎜
θ 23 = arccos ⎜
It is obvious that only in a specially designed curvilinear coordinate systems can the covariant vectors be mutually perpendicular. Such systems are called orthogonal. In this case the transformed region is rectangular. General curvilinear coordinate systems are not orthogonal in many applications. In such systems the transformed region where the curvilinear coordinates are independent variables can be thought of as being rectangular, and can be treated as such from a coding standpoint by formation of the finite difference
Appendix 2 Basics of the coordinate transformation theory
671
equations and in the solution thereof. The problem is thus much simpler in the transformed field, since the boundaries here are all thought of as rectangular. Orthogonality: In an orthogonal coordinate system: a) the two types of base vectors are parallel and b) three base vectors on each type are mutually perpendicular. The consequence is that the off diagonal terms of covariant metric tensor are zeros. Surface area increment: Consider a parallelepiped with finite sizes a1dξ , a 2 dη , and a3 dζ . The areas of the surfaces defined with constant curvilinear coordinates are
ξ = const ,
dS1 = ( a 2 × a3 ) dη d ζ ,
(43)
η = const ,
dS 2 = ( a3 × a1 ) dξ dζ ,
(44)
ζ = const ,
dS3 = ( a1 × a 2 ) dξ dη ,
(45)
respectively. Using the vector identity
( A × B ) ⋅ ( C × D ) = ( A ⋅ C )( B ⋅ D ) − ( A ⋅ D )( B ⋅ C )
(46)
for C = A and D = B
( A × B ) ⋅ ( A × B ) = ( A ⋅ A )( B ⋅ B ) − ( A ⋅ B ) ,
(47)
( A ⋅ A )( B ⋅ B ) − ( A ⋅ B )
(48)
2
or A×B =
2
the magnitude of the increment of the surface area is
ξ = const , dS 1 = a 2 × a3 dη dζ =
( a 2 ⋅ a 2 )( a3 ⋅ a3 ) − ( a 2 ⋅ a3 )
2
2 dη dζ = g 22 g33 − g 23 dη d ζ ,
(49)
η = const , dS 2 = a1 × a3 dξ dζ =
( a1 ⋅ a1 )( a3 ⋅ a3 ) − ( a3 ⋅ a1 )
2
2 dξ dζ = g11 g 33 − g 31 dξ dζ ,
(50)
ζ = const , dS 3 = a1 × a 2 dξ dη =
( a1 ⋅ a1 )( a2 ⋅ a2 ) − ( a1 ⋅ a2 )
2
dξ dη = g11 g 22 − g122 dξ dη .
(51)
672
Appendix 2 Basics of the coordinate transformation theory
Infinite volume in curvilinear coordinate systems: An infinitesimal parallelepiped formed by the vectors a1dξ , a 2 dη , and a3 dζ has a infinitesimal volume computed using the box product rule dV = ( a1dξ × a 2 dη ) ⋅ a3 dζ = ( a1 × a 2 ) ⋅ a3 dξ dη dζ
a11
a12
= a21
a22
a31
a32
∂x ∂ξ a13 ∂y a23 dξ dη dζ = ∂ξ a33 ∂z ∂ξ
∂x ∂η ∂y ∂η ∂z ∂η
∂x ∂ζ ∂y dξ dη dζ , ∂ζ ∂z ∂ζ
(52)
or using the Jacobian determinant or Jacobian of the coordinate transformation x = f (ξ ,η , ζ ) , y = g (ξ ,η , ζ ) , z = h (ξ ,η , ζ ) we have dV = J (ξ ,η , ζ ) dξ dη dζ .
(53)
Another notation of the Jacobian is g = J (ξ ,η , ζ ) = ( a1 × a 2 ) ⋅ a3 = ( a 2 × a3 ) ⋅ a1 = ( a3 × a1 ) ⋅ a 2 ,
(54)
so that the volume increment can also be written as dV = g dξ dη dζ .
(55)
Note the cyclic permutation of the subscripts inside the brackets in Eq. (54). In finite difference form we have ΔV = g ΔξΔηΔζ .
(56)
Selecting strict equidistant discretization into the transformed space with Δξ = 1 , Δη = 1 , Δζ = 1 we have a visualization of the meaning of the Jacobian, namely g = ΔV .
(57)
Some authors are using this approach to compute the Jacobian from the volume of the computational cell provided the discretization of the computational domain is strictly equidistant with steps of unity in all directions. The real volume of the cell is computed in this case from the Cartesian coordinates of the vertices of the real cell as will be shown later.
Appendix 2 Basics of the coordinate transformation theory
673
The Jacobian of the coordinate transformation in terms of dot products of the covariant vectors: The square of the Jacobian g = J (ξ ,η , ζ ) = ⎡⎣a1 ⋅ ( a 2 × a3 ) ⎤⎦ 2
2
(58)
can be rewritten as a function of the dot products of the components of the Jacobian determinant as follows. Using the vector identities
( A × B ) ⋅ ( C × D ) = ( A ⋅ C )( B ⋅ D ) − ( A ⋅ D )( B ⋅ C )
(59)
for C = A and D = B results in
( A × B ) ⋅ ( A × B ) = ( A ⋅ A )( B ⋅ B ) − ( A ⋅ B )
2
(60)
or
(A ⋅ B)
2
= ( A ⋅ A )( B ⋅ B ) − ( A × B ) ⋅ ( A × B ) .
(61)
With this result we have ⎡⎣a1 ⋅ ( a 2 × a3 ) ⎤⎦ = ( a1 ⋅ a1 )( a 2 × a3 ) ⋅ ( a 2 × a3 ) − ⎡⎣a1 × ( a 2 × a3 ) ⎤⎦ ⋅ ⎡⎣a1 × ( a 2 × a3 ) ⎤⎦ . (62) 2
Using the same identity Eq. (59)
( a2 × a3 ) ⋅ ( a2 × a3 ) = ( a 2 ⋅ a 2 )( a3 ⋅ a3 ) − ( a 2 ⋅ a3 )
2
(63)
and by the vector identity A × ( B × C) = ( A ⋅ C) B − ( A ⋅ B ) C
(64)
we have a1 × ( a 2 × a3 ) = ( a1 ⋅ a3 ) a 2 − ( a1 ⋅ a 2 ) a3 .
(65)
Finally we obtain 2 2 ⎡⎣a1 ⋅ ( a 2 × a3 ) ⎤⎦ = ( a1 ⋅ a1 ) ⎡( a 2 ⋅ a 2 )( a3 ⋅ a3 ) − ( a2 ⋅ a3 ) ⎤ − ( a1 ⋅ a3 ) a2 − ( a1 ⋅ a2 ) a3 ⎣ ⎦
2
= ( a1 ⋅ a1 )( a 2 ⋅ a 2 )( a3 ⋅ a3 ) − ( a1 ⋅ a1 )( a 2 ⋅ a3 ) − ( a 2 ⋅ a 2 )( a1 ⋅ a3 ) − ( a3 ⋅ a3 )( a1 ⋅ a 2 ) 2
2
+2 ( a1 ⋅ a 2 )( a3 ⋅ a 2 )( a1 ⋅ a3 ) a1 ⋅ a1
a1 ⋅ a 2
= a 2 ⋅ a1 a 2 ⋅ a 2 a3 ⋅ a1
(
a3 ⋅ a 2
a1 ⋅ a3 a 2 ⋅ a3 = gij a3 ⋅ a3
)
(
)
(
)
2 = g11 g 22 g33 − g 23 − g12 g 21 g33 − g 23 g31 + g13 g 21 g32 − g 22 g31 ,
which is the determinant of the covariant symmetric tensor.
(66)
2
674
Appendix 2 Basics of the coordinate transformation theory
Relation between the partial derivatives with respect to Cartesian and curvilinear coordinates: Partial derivatives with respect to Cartesian coordinates are related to partial derivatives with respect to the curvilinear coordinates by the chain rule. If ϕ ( x, y, z ) is a scalar-valued function of the main variables x, y, z , through the three intermediate variables ξ ,η , ζ , then ∂ϕ ∂ϕ ∂ξ ∂ϕ ∂η ∂ϕ ∂ζ ∂ϕ 11 ∂ϕ 21 ∂ϕ 31 = + + = a + a + a , ∂x ∂ξ ∂x ∂η ∂x ∂ζ ∂x ∂ξ ∂η ∂ζ
(67)
∂ϕ ∂ϕ ∂ξ ∂ϕ ∂η ∂ϕ ∂ζ ∂ϕ 12 ∂ϕ 22 ∂ϕ 32 = + + = a + a + a , ∂y ∂ξ ∂y ∂η ∂y ∂ζ ∂y ∂ξ ∂η ∂ζ
(68)
∂ϕ ∂ϕ ∂ξ ∂ϕ ∂η ∂ϕ ∂ζ ∂ϕ 13 ∂ϕ 23 ∂ϕ 33 = + + = a + a + a , ∂z ∂ξ ∂z ∂η ∂z ∂ζ ∂z ∂ξ ∂η ∂ζ
(69)
with the second superscript indicating the Cartesian component of the contravariant vectors. If V ( x, y , z ) = u ( x , y , z ) i + v ( x , y , z ) j + v ( x , y , z ) k
is a vector-valued function of the main variables x, y, z , through the three intermediate variables ξ ,η , ζ , then ∂V ⎛ ∂u 11 ∂u 21 ∂u 31 ⎞ ⎛ ∂v 11 ∂v 21 ∂v 31 ⎞ =⎜ a + a + a ⎟i + ⎜ a + a + a ⎟j ∂η ∂ζ ∂η ∂ζ ∂x ⎝ ∂ξ ⎠ ⎝ ∂ξ ⎠ ⎛ ∂w 11 ∂w 21 ∂w 31 ⎞ +⎜ a + a + a ⎟k, (70) ∂η ∂ζ ⎝ ∂ξ ⎠ ∂V ⎛ ∂u 12 ∂u 22 ∂u 32 ⎞ ⎛ ∂v 12 ∂v 22 ∂v 32 ⎞ =⎜ a + a + a ⎟i + ⎜ a + a + a ⎟j ∂η ∂ζ ∂η ∂ζ ∂y ⎝ ∂ξ ⎠ ⎝ ∂ξ ⎠ ⎛ ∂w 12 ∂w 22 ∂w 32 ⎞ +⎜ a + a + a ⎟k, (71) ∂η ∂ζ ⎝ ∂ξ ⎠ ∂V ⎛ ∂u 13 ∂u 23 ∂u 33 ⎞ ⎛ ∂v 13 ∂v 23 ∂v 33 ⎞ =⎜ a + a + a ⎟i +⎜ a + a + a ⎟j ∂η ∂ζ ∂η ∂ζ ∂z ⎝ ∂ξ ⎠ ⎝ ∂ξ ⎠ ⎛ ∂w 13 ∂w 23 ∂w 33 ⎞ +⎜ a + a + a ⎟ k. (72) ∂η ∂ζ ⎝ ∂ξ ⎠
The divergence theorem (Gauss-Ostrogradskii): The divergence theorem says that under suitable conditions the outward flux of a vector field across a closed surface (oriented outward) equals the triple integral of the divergence of the field over the region enclosed by the surface. The flux of vector F = Mi + Nj + Pk across a closed oriented surface S in the direction of the surface’s outward unit
Appendix 2 Basics of the coordinate transformation theory
675
normal field n equals the integral of ∇ ⋅ F over the region D enclosed by the surface:
∫∫ F ⋅ ndσ =∫∫∫ ∇ ⋅ FdV . S
(73)
D
Volume of space enveloped by a closed surface: Consider in a space closed surface S described by the position vector r(x,y,z). The volume inside the surface can be expressed in terms of the area integral over its boundary using the divergence theorem ⎛ ∂
∂
∂
⎞
∫∫ r ⋅ ndσ = ∫∫∫ ∇ ⋅ rdV = ∫∫∫ ⎜⎝ ∂x i + ∂y j + ∂z k ⎠⎟ ⋅ ( xi + yj + zk ) dV S
D
D
⎛ ∂x ∂y ∂z ⎞ = ∫∫∫ ⎜ + + ⎟ dV = 3∫∫∫ dV = 3D , ∂x ∂y ∂z ⎠ D ⎝ D
(74)
or D=
1 r ⋅ nd σ . 3 ∫∫ S
(75)
A practical application of this relation is the computation of the volume enclosed in plane elements with area S m , outwards unit normal vectors n m , and position vector of the centroids of the faces rm , D=
1 ∑ Smn m ⋅ rm . 3 m
(76)
The divergence theorem for a curvilinear coordinate system: Consider a differential element of volume D bounded by six faces lying on coordinate surfaces, as shown in Fig. A2.3. Divergence: Applying the Divergence Theorem we obtain
∫∫∫ ( ∇ ⋅ F )
g d ξ dη dζ
D
= ∫∫ F ⋅ ( a 2 × a3 ) dη dζ − ∫∫ F ⋅ ( a 2 × a3 ) dη dζ S1
S2
+ ∫∫ F ⋅ ( a3 × a1 ) d ξ dζ − ∫∫ F ⋅ ( a3 × a1 ) d ξ dζ S3
S4
+ ∫∫ F ⋅ ( a1 × a 2 ) d ξ dη − ∫∫ F ⋅ ( a1 × a 2 ) d ξ dη . S5
(77)
S6
Dividing by D and letting D approach zero we obtain the so called conservative expression for the divergence
676
Appendix 2 Basics of the coordinate transformation theory
∇⋅F =
⎫ 1 ⎧∂ ∂ ∂ ⎡⎣( a3 × a1 ) ⋅ F ⎤⎦ + ⎡⎣( a1 × a 2 ) ⋅ F ⎤⎦ ⎬ . ⎨ ⎡⎣( a 2 × a3 ) ⋅ F ⎤⎦ + ∂η ∂ζ g ⎩ ∂ξ ⎭ (78)
Having in mind that a 2 and a3 are independent of ξ , and a1 and a3 are independent of η , and a1 and a 2 are independent of ζ results in the so called first fundamental metric identity, Peyret (1996), ∂ ∂ ∂ ( a2 × a3 ) + ( a3 × a1 ) + ( a1 × a2 ) = 0 . ∂ξ ∂η ∂ζ
(79)
The so called non-conservative expression for the divergence is then ∇⋅F =
1 ⎡ ∂F ∂F ∂F ⎤ + ( a3 × a1 ) ⋅ + ( a1 × a 2 ) ⋅ ⎢( a 2 × a 3 ) ⋅ ⎥. ∂ξ ∂η ∂ζ ⎦ g ⎣
(80)
Curl: The Divergence Theorem is valid if the dot product is replaced by a cross product. The conservative form of the curl is ∇×F =
⎫ 1 ⎧ ∂ ∂ ∂ ⎡⎣( a3 × a1 ) × F ⎤⎦ + ⎡⎣( a1 × a 2 ) × F ⎤⎦ ⎬ , ⎨ ⎡⎣( a 2 × a3 ) × F ⎤⎦ + ∂ ξ ∂ η ∂ ζ g ⎩ ⎭ (81)
and the non-conservative form ∇×F =
1 ⎡ ∂F ∂F ∂F ⎤ + ( a3 × a1 ) × + ( a1 × a 2 ) × ⎢( a 2 × a 3 ) × ⎥. ∂ξ ∂η ∂ζ ⎦ g ⎣
(82)
Gradient: The Divergence Theorem is valid if the tensor or the vector F is replaced by a scalar ϕ . The conservative form of the gradient vector (gradient) of the differentiable function ϕ is ∇ϕ =
⎫ 1 ⎧ ∂ ∂ ∂ ⎡( a3 × a1 ) ϕ ⎤⎦ + ⎡( a1 × a 2 ) ϕ ⎤⎦ ⎬ ,(83) ⎨ ⎡( a 2 × a3 ) ϕ ⎤⎦ + ∂η ⎣ ∂ζ ⎣ g ⎩ ∂ξ ⎣ ⎭
and the non-conservative form ∇ϕ =
1 ⎡ ∂ϕ ∂ϕ ∂ϕ ⎤ + ( a3 × a1 ) + ( a1 × a 2 ) ⎢( a 2 × a 3 ) ⎥. ∂ξ ∂η ∂ζ ⎦ g ⎣
(84)
Laplacian: The expression for the Laplacian follows from the expression for the divergence and gradient as follows: The conservative form is ∇ 2ϕ = ∇ ⋅ ( ∇ϕ )
Appendix 2 Basics of the coordinate transformation theory
677
⎡ ∂ ⎧∂ ⎫⎤ ⎢ ⎪ ∂ξ ⎡⎣( a 2 × a3 ) ϕ ⎤⎦ + ∂η ⎡⎣( a3 × a1 ) ϕ ⎤⎦ ⎪⎥ ⎪⎪⎥ 1 ∂ ⎢ 1 ⎪⎪ ⎢( a 2 × a 3 ) ⋅ = ⎨ ⎬⎥ g ∂ξ ⎢ g ⎪ ⎪⎥ ∂ ⎢ ⎪+ ⎪⎥ ⎡⎣( a1 × a 2 ) ϕ ⎤⎦ ⎢ ∂ ζ ⎪ ⎪⎭⎥⎦ ⎩ ⎣ ⎡ ∂ ⎧ ∂ ⎫⎤ ⎢ ⎪ ∂ξ ⎡⎣( a 2 × a3 ) ϕ ⎤⎦ + ∂η ⎡⎣( a3 × a1 ) ϕ ⎤⎦ ⎪⎥ 1 ∂ ⎢ 1 ⎪⎪ ⎪⎪⎥ ⎢( a1 × a3 ) ⋅ + ⎨ ⎬⎥ g ∂η ⎢ g ⎪ ⎪⎥ ∂ ⎢ ⎪+ ⎪⎥ ⎡⎣( a1 × a 2 ) ϕ ⎤⎦ ⎢ ∂ ζ ⎪ ⎪⎭⎥⎦ ⎩ ⎣ ⎡ ∂ ⎧ ∂ ⎫⎤ ⎢ ⎪ ∂ξ ⎡⎣( a 2 × a3 ) ϕ ⎤⎦ + ∂η ⎡⎣( a3 × a1 ) ϕ ⎤⎦ ⎪⎥ ⎪⎪⎥ 1 ∂ ⎢ 1 ⎪⎪ ⎢( a1 × a 2 ) ⋅ + ⎨ ⎬⎥ . g ∂ζ ⎢ g ⎪ ⎪⎥ ∂ ⎢ ⎪+ ⎪⎥ ⎡⎣( a1 × a 2 ) ϕ ⎤⎦ ⎢ ⎩⎪ ∂ζ ⎭⎪⎥⎦ ⎣
(85)
The non-conservative form is ∇ 2ϕ = ∇ ⋅ ( ∇ϕ ) ⎧ ∂ ⎡ 1 ⎡ ∂ϕ ∂ϕ ∂ϕ ⎤ ⎤ ⎫ ⎪ ⎢( a 2 × a 3 ) ⋅ + ( a3 × a1 ) + ( a1 × a 2 ) ⎢( a 2 × a 3 ) ⎥⎥ ⎪ ∂ξ ∂η ∂ζ ⎦ ⎥⎦ ⎪ g ⎣ ⎪ ∂ξ ⎢⎣ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎡ ⎤ 1 ⎪ ∂ 1 ⎡ ∂ϕ ∂ϕ ∂ϕ ⎤ ⎪ = + a × a ⋅ a × a + a × a + a × a ⎢( 1 3 ) ( 3 1) ( 1 2 ) ⎥⎥ ⎬ . ⎨ ⎢( 2 3 ) ∂ξ ∂ξ ∂ζ ⎦ ⎥⎦ ⎪ g ⎪ ∂η ⎢⎣ g ⎣ ⎪ ⎪ ⎪ ⎪ ⎪ ∂ ⎡ ⎤ 1 ⎡ ∂ϕ ∂ϕ ∂ϕ ⎤ ⎪⎪ ⎪+ a × a + a × a + a × a ⎢( a1 × a 2 ) ⋅ ( ) ( ) ( ) 3 1 1 2 ⎢ 2 3 ⎥⎥ ⎪ ∂ζ ⎣⎢ ∂ξ ∂ξ ∂ζ ⎦ ⎦⎥ ⎪ g ⎣ ⎩ ⎭ (86)
The above relations were used for the first time to derive conservative transformed equations in the gas dynamics by Vivand and Vinokur in 1974.
678
Appendix 2 Basics of the coordinate transformation theory
Contravariant metric tensor: It is very interesting to note that in the Laplacian the nine scalar products form a symmetric tensor ⎛ a1 ⋅ a1 a1 ⋅ a 2 ⎜ g ij = ⎜ a 2 ⋅ a1 a 2 ⋅ a 2 ⎜ a 3 ⋅ a1 a 3 ⋅ a 2 ⎝
a1 ⋅ a3 ⎞ ⎟ a 2 ⋅ a3 ⎟ , a 3 ⋅ a 3 ⎟⎠
(87)
called the contravariant metric tensor. Using the identity
( A × B ) ⋅ ( C × D ) = ( A ⋅ C )( B ⋅ D ) − ( A ⋅ D )( B ⋅ C )
(88)
we compute the six contravariant metric numbers as follows a1 ⋅ a1 = =
1 2 g 22 g33 − g 23 , g
(
a2 ⋅ a2 = =
(
(
(
)
(90)
1 1 2 ( a1 × a 2 ) ⋅ ( a1 × a 2 ) = ⎡⎣( a1 ⋅ a1 )( a 2 ⋅ a 2 ) − (a1 ⋅ a 2 ) ⎤⎦ g g
)
(91)
1 1 ( a 2 × a3 ) ⋅ ( a3 × a1 ) = ⎡⎣( a 2 ⋅ a3 )( a3 ⋅ a1 ) − ( a 2 ⋅ a1 )( a3 ⋅ a3 )⎤⎦ g g
)
(
)
(93)
1 1 ( a3 × a1 ) ⋅ ( a1 × a2 ) = ⎡⎣( a3 ⋅ a1 )( a1 ⋅ a 2 ) − ( a3 ⋅ a 2 )( a1 ⋅ a1 )⎤⎦ g g
1 g31 g12 − g32 g11 . g
(
(92)
1 1 ( a2 × a3 ) ⋅ ( a1 × a 2 ) = ⎡⎣( a 2 ⋅ a1 )( a3 ⋅ a 2 ) − ( a 2 ⋅ a 2 )( a3 ⋅ a1 ) ⎤⎦ g g
1 g 21 g32 − g 22 g31 , g
a 2 ⋅ a3 = =
1 1 2 ( a3 × a1 ) ⋅ ( a3 × a1 ) = ⎡⎣( a3 ⋅ a3 )( a1 ⋅ a1 ) − ( a3 ⋅ a1 ) ⎤⎦ g g
1 g 23 g31 − g 21 g33 , g
a1 ⋅ a3 = =
(89)
1 g11 g 22 − g122 , g
a1 ⋅ a 2 = =
)
1 2 g33 g11 − g31 , g
a3 ⋅ a3 = =
1 1 2 ( a2 × a3 ) ⋅ ( a2 × a3 ) = ⎡⎣( a2 ⋅ a2 )( a3 ⋅ a3 ) − ( a 2 ⋅ a3 ) ⎤⎦ g g
)
(94)
Appendix 2 Basics of the coordinate transformation theory
679
Computing the determinant of the contravariant metric tensor we realize that a1 ⋅ a1 a1 ⋅ a 2 a 2 ⋅ a1 a 2 ⋅ a 2 a3 ⋅ a1 a3 ⋅ a 2
a1 ⋅ a3 1 a 2 ⋅ a3 = . g a3 ⋅ a3
(95)
Relation between the contravariant and the covariant base vectors, dual vectors: The non-conservative form of the gradient vector of the differentiable function ϕ 1 ⎡ ∂ϕ ∂ϕ ∂ϕ ⎤ + ( a3 × a1 ) + ( a1 × a 2 ) ⎢( a 2 × a 3 ) ⎥ ∂ξ ∂η ∂ζ ⎦ g ⎣
∇ϕ =
(96)
is applied for all of the coordinates ξ ,η , ζ . Since the three coordinates are independent of each other their derivatives with respect to each other are zero. With this in mind we obtain ∇ξ =
1 ⎡ ∂ξ ∂ξ ∂ξ + ( a3 × a1 ) + ( a1 × a 2 ) ⎢( a 2 × a 3 ) ∂ξ ∂η ∂ζ g ⎣
⎤ 1 ⎥= g ⎦
( a 2 × a3 ) , (97)
∇η =
1 ⎡ ∂η ∂η ∂η ⎤ 1 + ( a3 × a1 ) + ( a1 × a 2 ) ⎢( a 2 × a 3 ) ⎥= ∂ξ ∂η ∂ζ ⎦ g ⎣ g
( a3 × a1 ) , (98)
∇ζ =
1 ⎡ ∂ζ ∂ζ ∂ζ + ( a3 × a1 ) + ( a1 × a 2 ) ⎢( a 2 × a 3 ) ∂ξ ∂η ∂ζ g ⎣
( a1 × a 2 ) , (99)
⎤ 1 ⎥= g ⎦
or simply a1 =
a2 =
a3 =
1 g 1 g 1 g
( a 2 × a3 ) ,
(100)
( a3 × a1 ) ,
(101)
( a1 × a2 ) .
(102)
The contravariant vectors are not unit vectors. This result allows us to simplify the expressions for the gradient, divergence, curl, Laplacian, etc. Note that with these results the first fundamental metric identity, ∂ ∂ ∂ ( a 2 × a3 ) + ( a3 × a1 ) + ( a1 × a 2 ) = 0 , ∂η ∂ξ ∂ζ
(103)
680
Appendix 2 Basics of the coordinate transformation theory
can be written as ∂ ∂ξ
(
)
g a1 +
∂ ∂η
(
)
g a2 +
∂ ∂ζ
(
)
g a3 = 0 .
(104)
It is very important that the grid generated for numerical integration fulfills the above identity strictly. Otherwise spurious numerical errors are introduced in the computational results. Note the useful relations between the covariant and the contravariant vectors a1 ⋅ a1 =
1 g
a1 ⋅ ( a 2 × a3 ) = 1 , 1
a1 ⋅ a 2 = 0 ,
a2 ⋅ a2 =
a1 ⋅ a3 = 0 ,
a 2 ⋅ a3 = 0 ,
g
a 2 ⋅ a1 = 0 ,
a 2 ⋅ ( a3 × a1 ) = 1, a3 ⋅ a3 =
1 g
a3 ⋅ a1 = 0 ,
(105-107)
a3 ⋅ a 2 = 0 ,
(108-110)
a3 ⋅ ( a1 × a 2 ) = 1.
(111-113)
i
These properties distinguish the couple of vectors ai and a . Such vectors are called dual. Using these properties Eqs. (33) and (35) can be easily proved. Approximating the contravariant base vectors: Consider a parallelepiped with finite sizes a1Δξ , a2 Δη , and a3 Δζ . The areas of the surfaces defined with constant curvilinear coordinates are
ξ = const ,
S1 = ( a 2 × a3 ) ΔηΔζ = g a1ΔηΔζ ,
(114)
η = const ,
S 2 = ( a3 × a1 ) ΔξΔζ = g a 2 ΔξΔζ ,
(115)
ζ = const ,
S3 = ( a1 × a 2 ) ΔξΔη = g a3 ΔξΔη ,
(116)
respectively. Solving with respect to the contravariant vectors gives a1 =
a2 =
a3 =
S1 g ΔηΔζ
S2 g ΔξΔζ
S3 g ΔξΔη
,
(117)
,
(118)
.
(119)
Appendix 2 Basics of the coordinate transformation theory
681
If one selects Δξ = 1 , Δη = 1 , and Δζ = 1 , the contravariant vectors are then a1 =
a2 =
a3 =
S1 g
S2 g
S3 g
,
(120)
,
(121)
.
(122)
The contravariant metric tensor is in this case g ij = ai ⋅ a j =
Si ⋅ S j . g
(123)
In a sense of the numerical construction of the computational grid the above formulas can be used for computation of the contravariant vectors for equidistant grids with steps of unity in all directions in the transformed space. Relations between the inverse metrics and the metrics of the coordinate transformation: Usually in practical problems the inverse metrics are computed analytically or numerically and then the metrics are computed using the following procedure i ∂ξ ∂ξ ∂ξ 1 1 ∂x a1 := i+ j+ k= ( a 2 × a3 ) = ∂x ∂y ∂z g g ∂η ∂x ∂ζ
j ∂y ∂η ∂y ∂ζ
k ∂z , ∂η ∂z ∂ζ
(124)
where the Cartesian components are ∂ξ 1 ⎛ ∂y ∂z ∂z ∂y ⎞ = − ⎜ ⎟, ∂x g ⎝ ∂η ∂ζ ∂η ∂ζ ⎠
(125)
∂ξ 1 ⎛ ∂x ∂z ∂z ∂x ⎞ =− − ⎜ ⎟, ∂y g ⎝ ∂η ∂ζ ∂η ∂ζ ⎠
(126)
∂ξ 1 ⎛ ∂x ∂y ∂y ∂x ⎞ = − ⎜ ⎟. ∂z g ⎝ ∂η ∂ζ ∂η ∂ζ ⎠
(127)
682
Appendix 2 Basics of the coordinate transformation theory
Similarly we have for the other vectors i ∂η ∂η ∂η 1 1 ∂x a 2 := i+ j+ k= ( a3 × a1 ) = ∂x ∂y ∂z g g ∂ζ ∂x ∂ξ
j ∂y ∂ζ ∂y ∂ξ
k ∂z ∂ζ ∂z ∂ξ
(128)
resulting in ∂η 1 ⎛ ∂y ∂z ∂z ∂y ⎞ = − ⎜ ⎟, ∂x g ⎝ ∂ζ ∂ξ ∂ζ ∂ξ ⎠
(129)
∂η 1 ⎛ ∂x ∂z ∂z ∂x ⎞ =− − ⎜ ⎟, ∂y g ⎝ ∂ζ ∂ξ ∂ζ ∂ξ ⎠
(130)
∂η 1 ⎛ ∂x ∂y ∂y ∂x ⎞ = − ⎜ ⎟, ∂z g ⎝ ∂ζ ∂ξ ∂ζ ∂ξ ⎠
(131)
and i ∂ζ ∂ζ ∂ζ 1 1 ∂x a3 := i+ j+ k= ( a1 × a 2 ) = ∂x ∂y ∂z g g ∂ξ ∂x ∂η
j ∂y ∂ξ ∂y ∂η
k ∂z ∂ξ ∂z ∂η
(132)
resulting in ∂ζ 1 ⎛ ∂y ∂z ∂z ∂y ⎞ = − ⎜ ⎟, ∂x g ⎝ ∂ξ ∂η ∂ξ ∂η ⎠
(133)
∂ζ 1 ⎛ ∂x ∂z ∂z ∂x ⎞ =− − ⎜ ⎟, ∂y g ⎝ ∂ξ ∂η ∂ξ ∂η ⎠
(134)
∂ζ 1 ⎛ ∂x ∂y ∂y ∂x ⎞ = − ⎜ ⎟. ∂z g ⎝ ∂ξ ∂η ∂ξ ∂η ⎠
(135)
A plausibility proof of this computation is obtained by comparing the partial derivatives with the components of Eq. (8).
Appendix 2 Basics of the coordinate transformation theory
683
Restatement of the conservative derivative operations: In what follows the covariant differential operators in conservative form are given. Divergence: ∇⋅F =
1 ⎡ ∂ ⎢ g ⎣ ∂ξ
(
)
g a1 ⋅ F +
∂ ∂η
(
)
∂ ∂ζ
g a2 ⋅ F +
(
⎤ g a3 ⋅ F ⎥ . ⎦
)
(136)
Curl: ∇×F =
1 ⎡ ∂ ⎢ g ⎣ ∂ξ
(
)
∂ ∂η
∂ ∂η
(
g a1 × F +
(
(
⎤ g a 3 × F ⎥ . (137) ⎦
)
∂ ∂ζ
(
⎤ g a3ϕ ⎥ . ⎦
g a2 × F +
)
Gradient: ∇ϕ =
1 ⎡ ∂ ⎢ g ⎣ ∂ξ
(
)
g a1ϕ +
)
g a 2ϕ +
∂ ∂ζ
)
(138)
From this expression the conservative expressions for the first derivatives result ∂ϕ 1 ⎡ ∂ = ⎢ ∂x g ⎣ ∂ξ
(
g a11ϕ +
∂ϕ 1 ⎡ ∂ = ⎢ ∂y g ⎣ ∂ξ
(
g a 21ϕ +
∂ϕ 1 ⎡ ∂ = ⎢ ∂z g ⎣ ∂ξ
(
g a 31ϕ +
)
∂ ∂ζ
(
⎤ g a13ϕ ⎥ , ⎦
(139)
)
∂ ∂ζ
(
⎤ g a 33ϕ ⎥ , ⎦
(140)
)
∂ ∂ζ
(
⎤ g a 33ϕ ⎥ . ⎦
(141)
)
∂ ∂η
(
g a12ϕ +
)
∂ ∂η
(
g a 22ϕ +
)
∂ ∂η
(
g a 32ϕ +
)
)
)
Using the first fundamental metric identity the above expressions reduce to the simple non-conservative form, which are also immediately obtained by the chain rule. If conservative integration schemes are designed the use of the conservative form is strongly recommended. Laplacian: ∇ 2ϕ = ∇ ⋅ ( ∇ϕ )
684
Appendix 2 Basics of the coordinate transformation theory
⎧ ∂ ⎪⎧ 1 ⎡ ∂ ⎤ ⎪⎫ ⎫ ∂ ∂ g a1ϕ + g a 2ϕ + g a3ϕ ⎥ ⎬ ⎪ ⎪ ⎨a ⋅ ⎢ ∂η ∂ζ ⎪ ∂ξ ⎩⎪ ⎣ ∂ξ ⎦ ⎭⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎫ ⎤ 1 ⎪⎪ ∂ ⎪⎧ 2 ⎡ ∂ ∂ ∂ ⎪⎪ = g a1ϕ + g a 2ϕ + g a3ϕ ⎥ ⎬⎬ . ⎨+ ⎨a ⋅ ⎢ ∂η ∂ζ g ⎪ ∂η ⎪⎩ ⎣ ∂ξ ⎦ ⎪⎭ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ∂ ⎧⎪ 3 ⎡ ∂ ⎫ ⎤ ∂ ∂ ⎪⎪ g a1ϕ + g a 2ϕ + g a3ϕ ⎥ ⎬ ⎪ ⎪+ ⎨a ⋅ ⎢ ∂η ∂ζ ⎦ ⎭⎪⎭⎪ ⎩⎪ ∂ζ ⎩⎪ ⎣ ∂ξ
(
)
(
)
(
)
(
)
(
)
(
)
(
)
(
)
(
)
(142) Diffusion Laplacian: Usually the Laplacian appears in the diffusion terms of the conservation equations in the form ⎧ ∂ ⎪⎧ 1 ⎡ ∂ ⎤ ⎪⎫ ⎫ ∂ ∂ g a1ϕ + g a 2ϕ + g a3ϕ ⎥ ⎬ ⎪ ⎪ ⎨a ⋅ λ ⎢ ∂η ∂ζ ⎪ ∂ξ ⎪⎩ ⎣ ∂ξ ⎦ ⎪⎭ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎧ ⎫ ⎤ ⎪⎪⎪ 1 ⎪ ∂ ⎪ 2 ⎡ ∂ ∂ ∂ ∇ ⋅ ( λ∇ϕ ) = g a1ϕ + g a 2ϕ + g a3ϕ ⎥ ⎬⎬ , ⎨+ ⎨a ⋅ λ ⎢ ∂η ∂ζ g ⎪ ∂η ⎩⎪ ⎣ ∂ξ ⎦ ⎭⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ∂ ⎪⎧ 3 ⎡ ∂ ⎤ ⎪⎫⎪ ∂ ∂ 1 2 3 g aϕ + g aϕ + g a ϕ ⎥ ⎬⎪ ⎪+ ⎨a ⋅ λ ⎢ ∂η ∂ζ ⎪⎩ ∂ζ ⎩⎪ ⎣ ∂ξ ⎦ ⎭⎪⎪⎭
(
)
(
)
(
)
(
)
(
)
(
)
(
)
(
)
(
)
(143) where λ is a scalar-valued function of the local flow parameters. Restatement of the non-conservative derivative operations: In what follows the covariant differential operators in non-conservative form are given. Divergence: ∇ ⋅ F = a1 ⋅ Curl: ∇ × F = a1 ×
∂F ∂F ∂F + a2 ⋅ + a3 ⋅ . ∂ξ ∂η ∂ζ
(144)
∂F ∂F ∂F + a2 × + a3 × . ∂ξ ∂η ∂ζ
(145)
∂ϕ ∂ϕ ∂ϕ + a2 + a3 . ∂ξ ∂η ∂ζ
(146)
Gradient: ∇ϕ = a1
Appendix 2 Basics of the coordinate transformation theory
685
From this expression the non-conservative expressions for the first derivatives result ∂ϕ ∂ϕ ∂ϕ ∂ϕ = a11 + a 21 + a 31 , ∂x ∂ξ ∂η ∂ζ
(147)
∂ϕ ∂ϕ ∂ϕ ∂ϕ = a12 + a 22 + a 32 , ∂y ∂ξ ∂η ∂ζ
(148)
∂ϕ ∂ϕ ∂ϕ ∂ϕ = a13 + a 23 + a 33 , ∂z ∂ξ ∂η ∂ζ
(149)
with the second superscript indicating the Cartesian components of the contravariant vectors. Laplacian: ⎧ ∂ ⎡ ⎪ ⎢ ⎪ ∂ξ ⎢⎣ ⎪ ⎪ ⎪⎪ ∂ 1 ∇ 2ϕ = ∇ ⋅ ( ∇ϕ ) = ⎨+ g ⎪ ∂η ⎪ ⎪ ⎪ ∂ ⎪+ ⎩⎪ ∂ζ ⎧ ∂ ⎡ ⎪ ⎢ ⎪ ∂ξ ⎣⎢ ⎪ ⎪ 1 ⎪⎪ ∂ = ⎨+ g ⎪ ∂η ⎪ ⎪ ⎪ ∂ ⎪+ ⎪⎩ ∂ζ
⎛ ∂ϕ ∂ϕ ∂ϕ ⎞ ⎤ ⎫ g a1 ⋅ ⎜ a1 + a2 + a3 ⎟⎥ ⎪ ∂η ∂ζ ⎠ ⎥⎦ ⎪ ⎝ ∂ξ ⎪ ⎪ ⎪⎪ ⎡ ⎤ ⎛ ⎞ ∂ ϕ ∂ ϕ ∂ ϕ 2 1 2 3 g a ⋅ a + a + a ⎢ ⎜ ⎟⎥ ⎬ ∂η ∂ζ ⎠ ⎥⎦ ⎪ ⎢⎣ ⎝ ∂ξ ⎪ ⎪ ⎡ ⎤⎪ 3 ⎛ 1 ∂ϕ 2 ∂ϕ 3 ∂ϕ ⎞ +a +a ⎢ g a ⋅⎜a ⎟⎥⎪ ∂η ∂ζ ⎠ ⎥⎪ ⎢⎣ ⎝ ∂ξ ⎦⎭
⎫ ⎪ ⎪ ⎪ ⎪ ⎪⎪ ⎤ ⎞ ∂ ϕ ∂ ϕ . + g 22 + g 23 ⎟⎥ ⎬ ∂η ∂ζ ⎠ ⎥⎦ ⎪ ⎪ ⎪ ⎪ ⎤ ⎞ ∂ ϕ ∂ ϕ + g 32 + g 33 ⎟⎥ ⎪ ∂η ∂ζ ⎠ ⎥⎪ ⎦⎭
⎛ ∂ϕ ∂ϕ ∂ϕ ⎞ ⎤ g ⎜ g 11 + g 12 + g 13 ⎟⎥ ∂ξ ∂η ∂ζ ⎠ ⎥⎦ ⎝ ⎡ ⎛ 21 ∂ϕ ⎢ g ⎜g ∂ξ ⎢⎣ ⎝ ⎡ ⎛ 31 ∂ϕ ⎢ g ⎜g ∂ξ ⎢⎣ ⎝
(150)
686
Appendix 2 Basics of the coordinate transformation theory
Diffusion Laplacian: ⎧∂ ⎡ ⎪ ⎢ ⎪ ∂ξ ⎢⎣ ⎪ ⎪ 1 ⎪⎪ ∂ ∇ ⋅ ( λ∇ϕ ) = ⎨+ g ⎪ ∂η ⎪ ⎪ ⎪ ∂ ⎪+ ⎪⎩ ∂ζ ⎧ ∂ ⎡ ⎪ ⎢ ⎪ ∂ξ ⎣⎢ ⎪ ⎪ 1 ⎪⎪ ∂ = ⎨+ g ⎪ ∂η ⎪ ⎪ ⎪ ∂ ⎪+ ⎪⎩ ∂ζ
⎛ ∂ϕ ∂ϕ ∂ϕ ⎞ ⎤ ⎫ g a1 ⋅ λ ⎜ a1 + a2 + a3 ⎟⎥ ⎪ ∂η ∂ζ ⎠ ⎥⎦ ⎪ ⎝ ∂ξ ⎪ ⎪ ⎡ ⎤ ⎛ 1 ∂ϕ ∂ϕ ∂ϕ ⎞ ⎪⎪ 2 + a2 + a3 ⎢ g a ⋅λ ⎜a ⎟⎥ ⎬ ∂η ∂ζ ⎠ ⎦⎥ ⎪ ⎝ ∂ξ ⎣⎢ ⎪ ⎪ ⎡ ⎤⎪ ⎛ 1 ∂ϕ 3 2 ∂ϕ 3 ∂ϕ ⎞ +a +a ⎢ g a ⋅λ ⎜a ⎟⎥ ⎪ ∂η ∂ζ ⎠ ⎥⎪ ⎝ ∂ξ ⎣⎢ ⎦⎭ ⎫ ⎪ ⎪ ⎪ ⎪ ⎤ ∂ϕ ∂ϕ ⎞ ⎪⎪ + g 22 + g 33 ⎟⎥ ⎬ . ∂η ∂ζ ⎠ ⎦⎥ ⎪ ⎪ ⎪ ⎪ ⎤ ⎞ ∂ ϕ ∂ ϕ + g 32 + g 33 ⎟⎥⎪ ∂η ∂ζ ⎠ ⎥⎪ ⎦⎭
⎛ ∂ϕ ∂ϕ ∂ϕ ⎞ ⎤ g λ ⎜ g 11 + g 12 + g 13 ⎟⎥ ∂ξ ∂η ∂ζ ⎠ ⎥⎦ ⎝ ⎡ ⎛ 21 ∂ϕ ⎢ g λ⎜g ∂ξ ⎝ ⎣⎢ ⎡ ⎛ 31 ∂ϕ ⎢ g λ⎜g ∂ξ ⎢⎣ ⎝
(151)
Divergence of a tensor: Consider the dyadic product of the vector V, VV, which is a second order tensor. The divergence of this tensor is then ∇ ⋅ ( VV ) = ⎡⎣∇ ⋅ ( Vu ) ⎤⎦ i + ⎡⎣∇ ⋅ ( Vv ) ⎤⎦ j + ⎡⎣∇ ⋅ ( Vw ) ⎤⎦ k .
(152)
In this case the already derived expressions for divergence of vectors can be used. Another example of the divergence of a tensor is the expression ⎡ ⎛ ∂V ⎞ ⎤ ⎡ ⎛ ∂V ⎞ ⎤ ⎡ ⎛ ∂V ⎞ ⎤ T ∇ ⋅ (η∇V ) = ⎢∇ ⋅ ⎜η ⎟ ⎥ j + ⎢∇ ⋅ ⎜ η ⎟ ⎥ i + ⎢∇ ⋅ ⎜ η ⎟⎥ k . ⎣ ⎝ ∂x ⎠ ⎦ ⎣ ⎝ ∂y ⎠ ⎦ ⎣ ⎝ ∂z ⎠ ⎦
(153)
Also in this case the already derived expressions for divergence of vectors can be used. Diffusion Laplacian of a tensor: Usually the diffusion Laplacian of a tensor appears in the diffusion terms of the momentum conservation equations in the form ∇ ⋅ (η∇V )
Appendix 2 Basics of the coordinate transformation theory
687
which is equivalent to ∇ ⋅ (η∇V ) = ⎡⎣∇ ⋅ (η∇u ) ⎤⎦ i + ⎡⎣∇ ⋅ (η∇v ) ⎤⎦ j + ⎡⎣∇ ⋅ (η∇w ) ⎤⎦ k .
(154)
In this case the already derived expressions for each component are used. Derivatives along the normal to the curvilinear coordinate surface: If ϕ (ξ ,η , ζ ) is a scalar-valued function of the main variables ξ ,η , ζ , through the three intermediate variables x, y , z , then the derivatives along the normal to the curvilinear coordinate surface are
ξ = const ,
∂ϕ a1 = 1 ⋅ ∇ϕ = e1 ⋅∇ϕ , * ∂ξ a
(155)
η = const ,
∂ϕ a2 = 2 ⋅ ∇ ϕ = e 2 ⋅ ∇ϕ , * ∂η a
(156)
ζ = const ,
∂ϕ a3 = 3 ⋅∇ϕ = e3 ⋅∇ϕ . * ∂ζ a
(157)
Derivatives along the tangent of the curvilinear coordinate lines: If ϕ (ξ ,η , ζ ) is a scalar-valued function of the main variables ξ ,η , ζ , through the three intermediate variables x, y, z , then ∂ϕ ∂ϕ ∂x ∂ϕ ∂y ∂ϕ ∂z ⎛ ∂ϕ ∂ϕ ∂ϕ ⎞ ⎛ ∂x ∂y ∂z ⎞ = + + =⎜ i+ j+ k ⎟⋅⎜ i+ j+ k⎟ ∂ξ ∂x ∂ξ ∂y ∂ξ ∂z ∂ξ ⎝ ∂x ∂y ∂z ⎠ ⎝ ∂ξ ∂ξ ∂ξ ⎠ ∂r = ( ∇ϕ ) ⋅ = a1 ⋅∇ϕ , (158) ∂ξ ∂ϕ ∂ϕ ∂x ∂ϕ ∂y ∂ϕ ∂z ⎛ ∂ϕ ∂ϕ ∂ϕ ⎞ ⎛ ∂x ∂y ∂z ⎞ = + + =⎜ i+ j+ k ⎟⋅⎜ i+ j+ k⎟ ∂η ∂x ∂η ∂y ∂η ∂z ∂η ⎝ ∂x ∂y ∂z ⎠ ⎝ ∂η ∂η ∂η ⎠ ∂r = ( ∇ϕ ) ⋅ = a 2 ⋅∇ϕ , (159) ∂η ∂ϕ ∂ϕ ∂x ∂ϕ ∂y ∂ϕ ∂z ⎛ ∂ϕ ∂ϕ ∂ϕ ⎞ ⎛ ∂x ∂y ∂z ⎞ = + + =⎜ i+ j+ k ⎟⋅⎜ i+ j+ k⎟ ∂ζ ∂x ∂ζ ∂y ∂ζ ∂z ∂ζ ⎝ ∂x ∂y ∂z ⎠ ⎝ ∂ζ ∂ζ ∂ζ ⎠ ∂r = ( ∇ϕ ) ⋅ = a 3 ⋅ ∇ϕ . (160) ∂ζ
These are the three directional derivatives of ϕ ( x, y, z ) along ξ ,η , ζ , respectively. Another quick check is obtained by replacing the non-conservative form of
688
Appendix 2 Basics of the coordinate transformation theory
the gradient in the transformed coordinate systems and using the dual properties of the co- and contravariant vectors. Derivatives along the normal to the coordinate lines and tangent to the curvilinear coordinate surface: The vector ai × ai is normal to the coordinate line on which ξ i varies and is also tangent to the coordinate surface on which ξ i is constant. Using the relations between the covariant and the contravariant vectors and the vector identity
( B × C ) × A = − ⎡⎣( A ⋅ C ) B − ( A ⋅ B ) C⎤⎦
(161)
we have 1 g
a1 × a1 = =−
1 g
( g13a 2 − g12a3 ) , 1 g
a2 × a2 = =−
1 g
(162) 1 ⎡( a 2 ⋅ a1 ) a3 − ( a 2 ⋅ a3 ) a1 ⎤⎦ g ⎣
( a3 × a1 ) × a2 = −
( g 21a3 − g 23a1 ), 1 g
a3 × a3 = =−
1 ⎡( a1 ⋅ a3 ) a 2 − ( a1 ⋅ a 2 ) a3 ⎤⎦ g ⎣
( a2 × a3 ) × a1 = −
1 g
(163) 1 ⎡( a3 ⋅ a 2 ) a1 − ( a3 ⋅ a1 ) a 2 ⎤⎦ g ⎣
( a1 × a2 ) × a3 = −
( g32a1 − g31a2 ).
(164)
The magnitude is then 2
a1 × a1 = =
(
)
1⎡ g13 g13 g 22 − g12 g32 − g12 g13 g32 − g12 g33 ⎤ , ⎦ g⎣
(
2
a2 × a2 = =
1 2 g13 g 22 − 2 g12 g13 g 32 + g122 g 33 g
)
(
1 2 2 g 21 g33 − 2 g 21 g 23 g31 + g 23 g11 g
(
)
)
1⎡ g 21 g 21 g33 − g 23 g31 − g 23 g 21 g31 − g 23 g11 ⎤ , ⎦ g⎣
(
2
a3 × a3 =
)
(
1 2 2 g 32 g11 − 2 g31 g 21 g32 + g31 g 22 g
(
(165)
)
)
(166)
Appendix 2 Basics of the coordinate transformation theory
=
1⎡ g 32 g 32 g11 − g 31 g 21 − g31 g 21 g32 − g31 g 22 ⎤ . ⎦ g⎣
(
)
(
)
689
(167)
Comparing with
(
)
(
)
(
2 g = g11 g 22 g33 − g 23 − g12 g 21 g33 − g 23 g31 + g13 g 21 g32 − g 22 g31
)
(168)
results in computationally cheaper expressions 1 2 g11 g 22 g 33 − g 23 − 1, g
(169)
2
1 g 22 g11 g33 − g312 − 1, g
(170)
2
1 2 g 33 g11 g 22 − g 21 − 1. g
(171)
2
a1 × a1 = a2 × a2 = a3 × a3 =
(
)
(
)
(
)
With this result the derivatives normal to the coordinate line on which ξ i varies and are also tangent to the coordinate surface on which ξ i is constant are a1 × a1 a × a1 1
a2 × a2 a2 × a2 a3 × a3 a × a3 3
⋅∇ϕ = −
⋅ ∇ϕ = −
⋅ ∇ϕ = −
1
(
)
2 g11 g 22 g33 − g 23 −g
1
(
)
2 g 22 g11 g33 − g31 −g
1
(
)
2 g 33 g11 g 22 − g 21 −g
( g13a 2 − g12a3 ) ⋅∇ϕ ,
(172)
( g21a3 − g23a1 ) ⋅∇ϕ ,
(173)
( g32a1 − g31a 2 ) ⋅ ∇ϕ .
(174)
Replacing the non-conservative form of the gradient in the transformed coordinate systems and using the dual properties of the co- and contravariant vectors results in a × a1 1
a1 × a1
a × a2 2
a2 × a2
⋅∇ϕ = −
⋅ ∇ϕ = −
g13
∂ϕ ∂ϕ − g12 ∂η ∂ζ
(
)
2 g11 g 22 g33 − g 23 −g
g 23
∂ϕ ∂ϕ − g 21 ∂ξ ∂ζ
(
)
2 g 22 g11 g33 − g31 −g
,
(175)
,
(176)
690
Appendix 2 Basics of the coordinate transformation theory
a3 × a3 a3 × a3
⋅∇ϕ = −
g 32
∂ϕ ∂ϕ − g 31 ∂ξ ∂η
(
)
2 g 33 g11 g 22 − g 21 −g
.
(177)
Moving coordinate system: Consider a coordinate transformation defined by x = f (ξ ,η , ζ ) ,
(178)
y = g (ξ , η , ζ ) ,
(179)
z = h (ξ , η , ζ ) ,
(180)
which moves with time in space. The velocity ⎛ ∂x ⎞ ⎛ ∂y ⎞ ⎛ ∂z ⎞ Vcs = ⎜ ⎟ i+⎜ ⎟ j+ ⎜ ⎟ k, ⎝ ∂τ ⎠η ,ξ ,ζ ⎝ ∂τ ⎠η ,ξ ,ζ ⎝ ∂τ ⎠η ,ξ ,ζ
(181)
is the grid point velocity. Note an interesting property of the time derivative of the covariant unit vectors ∂a1 ∂ ⎛ ∂r ⎞ ∂ ⎛ ∂r ⎞ ∂Vcs = , ⎜ ⎟= ⎜ ⎟= ∂τ ∂τ ⎝ ∂ξ ⎠ ∂ξ ⎝ ∂τ ⎠ ∂ξ
(182)
∂a 2 ∂ ⎛ ∂r ⎞ ∂ ⎛ ∂r ⎞ ∂Vcs = , ⎜ ⎟= ⎜ ⎟= ∂τ ∂τ ⎝ ∂η ⎠ ∂η ⎝ ∂τ ⎠ ∂η
(183)
∂a3 ∂ ⎛ ∂r = ⎜ ∂τ ∂τ ⎝ ∂ζ
(184)
⎞ ∂ ⎛ ∂r ⎞ ∂Vcs . ⎟= ⎜ ⎟= ⎠ ∂ζ ⎝ ∂τ ⎠ ∂ζ
If f = f (τ , x, y, z ) is a function of time and space using the chain rule we obtain df ⎛ ∂f ⎞ = dτ ⎜⎝ ∂τ ⎟⎠ x , y , z ⎡⎛ ∂f ⎞ ⎤ ⎡⎛ dx ⎞ ⎤ ⎛ ∂f ⎞ ⎛ ∂f ⎞ ⎛ dy ⎞ ⎛ dz ⎞ + ⎢ ⎜ ⎟ i + ⎜ ⎟ j + ⎜ ⎟ k ⎥ ⋅ ⎢⎜ i+⎜ j+⎜ k⎥ . ⎟ ⎟ ⎟ ⎝ ∂z ⎠t , x , y ⎦⎥ ⎣⎢⎝ dτ ⎠ξ ,η ,ζ ⎝ dτ ⎠ξ ,η ,ζ ⎝ dτ ⎠ξ ,η ,ζ ⎥⎦ ⎝ ∂y ⎠t , x , z ⎣⎢⎝ ∂x ⎠t , y , z (185)
Comparing the second part of the right hand side of the above equation with the definition of the gradient and of the grid point velocity we see that ⎛ df ⎞ ⎛ ∂f ⎞ =⎜ ⎟ + ( ∇f ) ⋅ Vcs , ⎜ dτ ⎟ ⎝ ⎠ξ ,η ,ζ ⎝ ∂τ ⎠ x , y , z
(186)
Appendix 2 Basics of the coordinate transformation theory
691
the absolute change with the time of a function f is the velocity with the respect to the stationary coordinate system xyz plus a component resulting from the movement of the coordinate system ξ ,η , ζ . Therefore ⎛ ∂f ⎞ ⎛ ∂f ⎞ =⎜ ⎟ − Vcs ⋅∇f . ⎜ ∂τ ⎟ ⎝ ⎠ x , y , z ⎝ ∂τ ⎠ξ ,η ,ζ
(187)
Here the time derivative ( df dt )ξ ,η ,ζ is understood to be at a fixed point in the transformed region. In a moving coordinate system the orientation of the unit covariant vectors will change. That is the elementary parallelepiped volume built by these vectors, the Jacobian, changes also. The derivative of the Jacobian of the coordinate transformation with time can be visualized by doting the change of the normal component to the covariant vector with the corresponding elementary coordinate surfaces formed by the other two vectors ⎛d g ⎜ ⎜ dτ ⎝
⎞ da da da = 1 ⋅ ( a 2 × a3 ) + 2 ⋅ ( a3 × a1 ) + 3 ⋅ ( a1 × a 2 ) ⎟ ⎟ d τ d τ dτ ⎠ξ ,η ,ζ
(188)
or d g
da da ⎛ da ⎞ = g ⎜ 1 ⋅ a1 + 2 ⋅ a 2 + 3 ⋅ a3 ⎟ , dτ dτ dτ ⎝ dτ ⎠
(189)
see Thompson, Warsi and Mastin (1985) p.130. Having in mind that ∂a1 ∂Vcs = , ∂τ ∂ξ
(190)
∂a 2 ∂Vcs = , ∂τ ∂η
(191)
∂a3 ∂Vcs = , ∂τ ∂ζ
(192)
results in ⎛ ∂V ∂V ∂V = g ⎜ a1 ⋅ cs + a 2 ⋅ cs + a3 ⋅ cs dτ ∂ξ ∂η ∂ζ ⎝
d g
⎞ ⎟. ⎠
(193)
Note some very interesting properties of the following expression ∂ ∂ξ
(
)
g a1 f ⋅ Vcs +
∂ ∂η
(
)
g a 2 f ⋅ Vcs +
∂ ∂ζ
(
g a3 f ⋅ Vcs
)
692
Appendix 2 Basics of the coordinate transformation theory
⎛ ∂V ∂V ∂V = f g ⎜ a1 ⋅ cs + a 2 ⋅ cs + a3 ⋅ cs ∂ξ ∂η ∂ζ ⎝ ⎡ ∂ ∂ ∂ +⎢ g a1 f + g a2 f + ∂η ∂ζ ⎣ ∂ξ
(
= f
)
⎡ ∂ +⎢ dτ ⎣ ∂ξ
d g
(
(
)
g a1 f +
∂ ∂η
⎞ ⎟ ⎠ ⎤ g a3 f ⎥ ⋅ Vcs ⎦
)
(
)
(
g a2 f +
)
∂ ∂ζ
⎤ g a3 f ⎥ ⋅ Vcs ⎦
(
)
(194)
Transformed fluid properties conservation equation: Consider f = f (τ , x, y, z ) being a function of time and space which is a controlled by the following conservation equation ⎛ ∂f ⎞ + ∇ ⋅ ( fV ) − ∇ ⋅ ( λ∇f ) = μ . ⎜ ∂τ ⎟ ⎝ ⎠ x, y, z
(195)
The convection-diffusion problem described by this equation is called isotropic. Using the expression for the time derivative, the conservative form of the divergence of a vector and gradient of a scalar in the transformed space, we obtain
(
∂ f g ∂τ
)
+
∂ ∂ξ
⎛ ⎡ ∂ 1 1 ⎜⎜ f g a ⋅ ( V − Vcs ) − a ⋅ λ ⎢ ⎣ ∂ξ ⎝
(
g a1 f +
+
⎡ ∂ ∂ ⎛ 2 2 ⎜⎜ f g a ⋅ ( V − Vcs ) − a ⋅ λ ⎢ ∂η ⎝ ⎣ ∂ξ
(
g a1 f +
⎛ ⎡ ∂ 3 3 ⎜⎜ f g a ⋅ ( V − Vcs ) − a ⋅ λ ⎢ ⎣ ∂ξ ⎝ = g μ.
(
g a1 f +
+
∂ ∂ζ
)
∂ ∂η
)
)
)
∂ ∂ζ
(
⎤⎞ g a3 f ⎥ ⎟ ⎟ ⎦⎠
)
∂ ∂ζ
(
⎤⎞ g a3 f ⎥ ⎟ ⎟ ⎦⎠
)
∂ ∂ζ
(
g a2 f +
∂ ∂η
(
g a2 f +
∂ ∂η
(
g a2 f +
)
)
⎤⎞ g a3 f ⎥ ⎟⎟ ⎦⎠ (196)
(
)
Using the non-conservative form of the Diffusion Laplacian results in a simpler form which is still conservative with respect to the spatial derivatives in the transformed coordinate system
Appendix 2 Basics of the coordinate transformation theory
(
∂ f g ∂τ
693
)
⎧⎪ ⎡ 1 ⎛ 1 1 ∂f ∂f ∂f ⎞ ⎤ ⎫⎪ + a1 ⋅ a 2 + a1 ⋅ a3 ⎨ g ⎢ fa ⋅ ( V − Vcs ) − λ ⎜ a ⋅ a ⎟⎥ ⎬ ∂ξ ∂η ∂ζ ⎠ ⎥⎦ ⎪⎭ ⎢⎣ ⎝ ⎩⎪
+
∂ ∂ξ
+
⎡ 2 ⎛ 2 1 ∂f ∂ ⎧⎪ ∂f ∂f ⎞ ⎤ ⎪⎫ + a2 ⋅ a2 + a 2 ⋅ a3 ⎨ g ⎢ fa ⋅ ( V − Vcs ) − λ ⎜ a ⋅ a ⎟⎥⎬ ∂η ⎩⎪ ∂ξ ∂η ∂ζ ⎠ ⎥⎦ ⎪⎭ ⎢⎣ ⎝
+
∂ ∂ζ
⎧⎪ ⎡ 3 ⎛ 3 1 ∂f ∂f ∂f ⎞ ⎤ ⎫⎪ + a3 ⋅ a 2 + a3 ⋅ a3 ⎨ g ⎢ fa ⋅ ( V − Vcs ) − λ ⎜ a ⋅ a ⎟⎥⎬ = g μ . ∂ξ ∂η ∂ζ ⎠ ⎥⎦ ⎪⎭ ⎢⎣ ⎝ ⎪⎩ (197)
The equation can be rewritten in more compact form using the relative contravariant velocity components in the transformed space defined as follows ⎛ a1 ⋅ ( V − Vcs ) ⎞ ⎛ V 1 ⎞ ⎜ ⎟ ⎜ ⎟ V = ⎜ a 2 ⋅ ( V − Vcs ) ⎟ = ⎜ V 2 ⎟ ⎜ a3 ⋅ ( V − V ) ⎟ ⎜ V 3 ⎟ cs ⎠ ⎝ ⎝ ⎠
(198)
and the terms of the inverse metrics matrices
(
∂ f g ∂τ
)
⎧⎪ ⎡ 1 ⎛ 11 ∂f ∂f ∂f ⎞ ⎤ ⎫⎪ + g 12 + g 13 ⎨ g ⎢ fV − λ ⎜ g ⎟⎥ ⎬ ∂ξ ∂η ∂ζ ⎠ ⎥⎦ ⎪⎭ ⎢⎣ ⎝ ⎩⎪
+
∂ ∂ξ
+
⎡ 2 ⎛ 21 ∂f ∂ ⎧⎪ ∂f ∂f ⎞ ⎤ ⎪⎫ + g 22 + g 23 ⎨ g ⎢ fV − λ ⎜ g ⎟⎥ ⎬ ∂η ⎩⎪ ∂ξ ∂η ∂ζ ⎠ ⎥⎦ ⎪⎭ ⎢⎣ ⎝
+
∂ ∂ζ
⎧⎪ ⎡ 3 ⎛ 31 ∂f ∂f ∂f ⎞ ⎤ ⎫⎪ + g 32 + g 33 ⎨ g ⎢ fV − λ ⎜ g ⎟⎥ ⎬ = g μ . ∂ξ ∂η ∂ζ ⎠ ⎥⎦ ⎪⎭ ⎢⎣ ⎝ ⎩⎪
(199)
We see a remarkable property of this equation. In the transformed coordinate system the property f g having convection-diffusion flux i-components
694
Appendix 2 Basics of the coordinate transformation theory
⎡ ⎛ ∂f ∂f ∂f ⎞ ⎤ Iˆi = g ⎢ fai ⋅ ( V − Vcs ) − λ ⎜ g i1 + gi2 + g i3 ⎟⎥ ∂ξ ∂η ∂ζ ⎠ ⎥⎦ ⎝ ⎣⎢
(200)
is conserved. The convective component can be visualized as g fai ⋅ ( V − Vcs ) = g f ai ei ⋅ ( V − Vcs ) = g f ai V n,i .
(201)
Here V n ,i = ei ⋅ ( V − Vcs )
(202)
is the relative velocity component normal to the surface defined by ξ i = const . Remember Eqs. (117-119) giving g ai =
Si , dξ j dξ k
(203)
or g a =
Si
i
dξ j dξ k
.
(204)
For equidistant discretization in the computational space we have g a i = S i = S i.
(205)
Therefore the convective flux can be presented in several ways g fai ⋅ ( V − Vcs ) = g f ai ei ⋅ ( V − Vcs ) = S iV n, i f = g V i f = S i ⋅ ( V − Vcs ) f .
(206) With this the conservation equation can be rewritten in terms of the normal velocity components
(
∂ f g ∂τ
)
+
∂ ∂ξ
⎡ 1 n ,1 ⎛ 11 ∂f ∂f ∂f ⎞ ⎤ + g 12 + g 13 ⎢S V f − λ g ⎜ g ⎟⎥ ∂ ξ ∂ η ∂ ζ ⎠ ⎥⎦ ⎝ ⎣⎢
+
∂ ∂η
⎡ 2 n ,2 ⎛ 21 ∂f ∂f ∂f ⎞ ⎤ + g 22 + g 23 ⎢S V f − λ g ⎜ g ⎟⎥ ∂ξ ∂η ∂ζ ⎠ ⎥⎦ ⎢⎣ ⎝
Appendix 2 Basics of the coordinate transformation theory
+
∂ ∂ζ
⎡ 3 n,3 ⎛ 31 ∂f ∂f ∂f ⎞ ⎤ + g 32 + g 33 ⎢S V f − λ g ⎜ g ⎟⎥ = g μ . ∂ξ ∂η ∂ζ ⎠ ⎥⎦ ⎝ ⎣⎢
695
(207)
The convection-diffusion problem in the transformed space is called anisotropic. Thus, the isotropic convection-diffusion problem in the physical space turns out to be anisotropic in the transformed space. Note that it is computationally more effective to store g ij g instead of g ij if g does not change with the time.
For an orthogonal transformed system the equation simplifies much more ⎤ ∂ ⎡ ⎤ ⎛ 1 ⎛ 2 ∂ ∂ ⎡ 11 ∂f ⎞ 22 ∂f ⎞ f g + ⎢ g ⎜ fV − λ g ⎢ g ⎜ fV − λ g ⎟⎥ + ⎟⎥ ∂τ ∂ξ ⎣ ∂ξ ⎠ ⎦ ∂η ⎣ ∂η ⎠ ⎦ ⎝ ⎝ ⎤ ⎛ 3 ∂ ⎡ 33 ∂f ⎞ + (208) ⎢ g ⎜ fV − λ g ⎟⎥ = g μ . ∂ζ ⎣ ∂ ζ ⎝ ⎠⎦
(
)
The convection-diffusion problem in an orthogonal transformed system is called orthotropic. Time dependent grid metric identity (second fundamental metric identity): Consider a conservation equation in the transformed space for f = const, and V = const without diffusion and mass sources ∂ g ∂ − ∂τ ∂ξ
(
g a1 ⋅ Vcs −
⎡ ∂ = −V ⋅ ⎢ ⎣ ∂ξ
(
g a1 −
)
)
∂ ∂η
∂ ∂η
(
(
)
g a 2 ⋅ Vcs −
)
g a2 −
∂ ∂ζ
(
∂ ∂ζ
g a3 ⋅ Vcs
)
⎤ g a3 ⎥ . ⎦
(
)
(209)
Using the first fundamental metric identity Thompson et al. (1985), p.159, obtained ∂ g ∂τ
−
∂ ∂ξ
(
)
g a1 ⋅ Vcs −
∂ ∂η
(
)
g a 2 ⋅ Vcs −
∂ ∂ζ
(
)
g a 3 ⋅ Vcs = 0. (210)
This is an important identity for the numerical analysis called sometimes the second fundamental metric identity. As pointed out by Thompson et al. (1985) this identity should be used to numerically determine updated values of the Jacobian,
g , instead of updating it directly from the new values of the Cartesian coordinates. In the later case spurious source terms will appear Thompson et al. (1985). Transformed total time derivative of a scalar: The total derivative of a scalar ϕ traveling in the Cartesian coordinate system with velocity V is defined with
696
Appendix 2 Basics of the coordinate transformation theory
dϕ ∂ϕ = + V ⋅∇ϕ . dτ ∂τ
(211)
Using the already derived relations we obtain dϕ ⎛ ∂ϕ ⎞ = + Vl ⋅∇ϕ dτ ⎜⎝ ∂τ ⎟⎠ x , y , z ∂ϕ ∂ϕ ∂ϕ ⎛ ∂ϕ ⎞ =⎜ + a1 ⋅ ( Vl − Vcs ) + a 2 ⋅ ( Vl − Vcs ) + a3 ⋅ ( Vl − Vcs ) ⎟ ∂ξ ∂η ∂ζ ⎝ ∂τ ⎠ξ ,η ,ζ ∂ϕ ∂ϕ ∂ϕ ⎛ ∂ϕ ⎞ =⎜ + Vl1 + Vl 2 + Vl 3 . ⎟ ∂ξ ∂η ∂ζ ⎝ ∂τ ⎠ξ ,η ,ζ
(212)
Examples of some frequently used curvilinear coordinate transformations: Next we apply the derivations from this Appendix to two well-known coordinate systems – the cylindrical and the spherical. If the derived formulas are applied correctly the well-known expressions will finally arise for the differential operators. It is recommended to go through these examples in order to understand better the transformation theory. Cylindrical coordinates: The transformation is defined analytically by x = ξ cos η ,
ξ = + x2 + y2 ,
(213,214)
y = ξ sin η ,
η = arctan ( y x ) ,
(215,216)
z =ζ,
ζ = z.
(217,218)
Here ξ is the radial coordinate, η is the azimuthal coordinate and the axial coordinate ζ is equivalent with z. An equidistant transformation in the computational space is defined by ⎡ ⎛ θ − θ min ⎞ ξ ⎤ x (ξi ,η j ) = ⎢ rmin + ( rmax − rmin ) i ⎥ cos ⎜ max η j ⎟, i jmax ⎣ max ⎦ ⎝ ⎠ ⎡ ξ ⎤ ⎛ θ − θ min ⎞ y (ξi ,η j ) = ⎢ rmin + ( rmax − rmin ) i ⎥ sin ⎜ max η j ⎟, imax ⎦ ⎝ jmax ⎣ ⎠ z − zmin z (ζ k ) = max ζk, kmax
where
Appendix 2 Basics of the coordinate transformation theory
697
ξi = 0,1, 2,..., imax , η j = 0,1, 2,..., jmax , ζ k = 0,1, 2,..., kmax. The inverse metrics of the coordinate transformation are then ⎛ ∂x ⎜ ⎜ ∂ξ ⎜ ∂y ⎜ ⎜ ∂ξ ⎜ ∂z ⎜⎜ ∂ξ ⎝
∂x ∂η ∂y ∂η ∂z ∂η
∂x ∂ζ ∂y ∂ζ ∂z ∂ζ
⎞ ⎟ ⎟ ⎛ cos η ⎟ ⎜ ⎟ = ⎜ sin η ⎟ ⎜ 0 ⎟ ⎝ ⎟⎟ ⎠
−ξ sin η 0 ⎞ ⎟ ξ cos η 0 ⎟ . 0 1 ⎟⎠
(219)
The Jacobian is then g =
∂x ∂ξ
= ξ.
⎛ ∂y ∂z ∂y ∂z ⎞ ∂x ⎛ ∂y ∂z ∂y ∂z ⎞ ∂x ⎛ ∂y ∂z ∂y ∂z ⎞ − − − ⎜ ⎟− ⎜ ⎟+ ⎜ ⎟ ⎝ ∂η ∂ζ ∂ζ ∂η ⎠ ∂η ⎝ ∂ξ ∂ζ ∂ζ ∂ξ ⎠ ∂ζ ⎝ ∂ξ ∂η ∂η ∂ξ ⎠ (220)
The metrics of the coordinate transformation are ⎛ ∂ξ ⎜ ⎜ ∂x ⎜ ∂η ⎜ ⎜ ∂x ⎜ ∂ζ ⎜⎜ ∂x ⎝
∂ξ ∂y ∂η ∂y ∂ζ ∂y
⎛ ∂y ∂z ∂y ∂z ∂ξ ⎞ − ⎟ ⎜ ∂z ⎟ ⎜ ∂η ∂ζ ∂ζ ∂η ∂η ⎟ 1 ⎜ ∂y ∂z ∂y ∂z − ⎟= ⎜ ∂z ⎟ g ⎜ ∂ζ ∂ξ ∂ξ ∂ζ ⎜ ∂y ∂z ∂y ∂z ∂ζ ⎟ ⎟ ⎟ ⎜⎜ ∂ξ ∂η − ∂η ∂ξ ∂z ⎠ ⎝
⎛ cosη ⎜ 1 = ⎜ − sin η ⎜ ξ ⎜⎜ 0 ⎝
∂x ∂z ∂x ∂z − ∂ζ ∂η ∂η ∂ζ ∂x ∂z ∂x ∂z − ∂ξ ∂ζ ∂ζ ∂ξ ∂x ∂z ∂x ∂z − ∂η ∂ξ ∂ξ ∂η
∂x ∂η ∂x ∂ζ ∂x ∂ξ
∂y ∂x ∂y ⎞ − ⎟ ∂ζ ∂ζ ∂η ⎟ ∂y ∂x ∂y ⎟ − ⎟ ∂ξ ∂ξ ∂ζ ⎟ ∂y ∂x ∂y ⎟ − ⎟ ∂η ∂η ∂ξ ⎟⎠
sin η
0⎞ ⎟ 1 cos η 0 ⎟ . ⎟ ξ ⎟ 0 1 ⎟⎠
(221)
The components of the three covariant base vectors of the curvilinear coordinate system are the columns of the matrix of the inverse metrics. They are given below together with the corresponding covariant unit vectors. a1 =
∂x ∂y ∂z i+ j+ k = cos η i + sin η j ∂ξ ∂ξ ∂ξ
a1 = 1
e1 = cosη i + sin η j ,
a2 =
∂x ∂y ∂z i+ j+ k = −ξ sin η i + ξ cos η j ∂η ∂η ∂η
a2 = ξ
e 2 = − sin η i + cosη j
a3 =
∂x ∂y ∂z i+ j+ k =k ∂ζ ∂ζ ∂ζ
a3 = 1
e3 = k
(222-230)
698
Appendix 2 Basics of the coordinate transformation theory
The components of the three contravariant base vectors of the curvilinear coordinate system are the rows of the matrix of the metrics. They are given below together with the corresponding contravariant unit vectors. a1 =
∂ξ ∂ξ ∂ξ i+ j+ k = cosη i + sin η j ∂x ∂y ∂z
a1 = 1
a2 =
∂η ∂η ∂η 1 1 i+ j+ k = − sin η i + cosη j ∂x ∂y ∂z ξ ξ
a2 =
a3 =
∂ζ ∂ζ ∂ζ i+ j+ k =k ∂x ∂y ∂z
a3 = 1
e1 = cosη i + sin η j 1
e2 = − sin η i + cos η j
ξ
e3 = k
(231-239) We see that a) the two types of base vectors are parallel and that b) three base vectors on each type are mutually perpendicular which means that the cylindrical coordinate system is orthogonal. The gradient of the vector is then 1 ⎡ ∂ ⎢ g ⎣ ∂ξ
∇⋅F =
= =
1 ⎡ ∂ ⎢ g ⎣ ∂ξ
(
(
)
g a1 ⋅ F +
)
g a1 e1 ⋅ F +
∂ ∂η
∂ ∂η
(
(
)
g a2 ⋅ F +
)
g a2 e2 ⋅ F +
1 ∂ 1 ∂ ∂ ξ Fξ + Fη + Fζ , ξ ∂ξ ξ ∂η ∂ζ
(
)
( )
∂ ∂ζ
( )
(
∂ ∂ζ
⎤ g a3 ⋅ F ⎥ , ⎦
)
(
⎤ g a1 e3 ⋅ F ⎥ , ⎦
)
(240)
where Fξ = e1 ⋅ F , Fη = e 2 ⋅ F , Fζ = e3 ⋅ F are the components of F normal to the coordinate surfaces defined by ξ = const , η = const , ζ = const , respectively. This is the expected form for cylindrical coordinates – compare for instance Bird, Stewart and Lightfood (1960), p. 739B. The covariant metric tensor is ⎛ a1 ⋅ a1 a1 ⋅ a 2 ⎜ gij = ⎜ a 2 ⋅ a1 a 2 ⋅ a 2 ⎜a ⋅a a ⋅a 3 2 ⎝ 3 1
a1 ⋅ a3 ⎞ ⎛ 1 0 ⎟ ⎜ a 2 ⋅ a3 ⎟ = ⎜ 0 ξ 2 a3 ⋅ a3 ⎟⎠ ⎜⎝ 0 0
0⎞ ⎟ 0⎟ , 1 ⎟⎠
(241)
which is a symmetric tensor. The angles between the covariant unit vectors are ⎛
θ12 = arccos ⎜
g12
⎜ g g ⎝ 11 22
⎞ π ⎟= , ⎟ 2 ⎠
(242)
Appendix 2 Basics of the coordinate transformation theory
θ13 = arccos ⎜
⎛
g13
⎞ π ⎟= , ⎜ g g ⎟ 2 ⎝ 11 33 ⎠ ⎛
g 23
(243)
⎞ π ⎟= . ⎜ g g ⎟ 2 22 33 ⎠ ⎝
θ 23 = arccos ⎜
699
(244)
The magnitude of the increment of the surface area is
ξ = const ,
2 dS 1 = g 22 g 33 − g 23 dη d ζ = ξ dη dζ ,
(245)
η = const ,
dS 2 = g11 g33 − g132 dξ dζ = dξ d ζ ,
(246)
ζ = const ,
dS 3 = g11 g 22 − g122 dξ dη = ξ d ξ dη .
(247)
The volume increment is then dV = g dξ dη dζ = ξ d ξ dη d ζ .
(248)
The elements of the contravariant metric tensor are ⎛ a1 ⋅ a1 a1 ⋅ a 2 ⎜ g ij = ⎜ a 2 ⋅ a1 a 2 ⋅ a 2 ⎜ a 3 ⋅ a1 a 3 ⋅ a 2 ⎝
a1 ⋅ a3 ⎞ ⎟ a 2 ⋅ a3 ⎟ a 3 ⋅ a 3 ⎟⎠
2 ⎛ g 22 g33 − g 23 1⎜ = ⎜ g 21 g33 − g 23 g31 g⎜ ⎝ g 21 g32 − g 22 g31
g 21 g33 − g 23 g31 g11 g33 − g132 g11 g32 − g12 g31
⎛1 0 g 21 g 32 − g 22 g 31 ⎞ ⎜ ⎟ 1 g11 g 32 − g12 g 31 ⎟ = ⎜ 0 2 ⎜ ξ g11 g 22 − g122 ⎟⎠ ⎜⎜ ⎝0 0
0⎞ ⎟ 0⎟ . ⎟ ⎟ 1 ⎟⎠ (249)
With this result we can compute the Diffusion Laplacian of a scalar for the orthogonal system ∇ ⋅ ( λ∇ϕ ) = =
⎤ 1 ⎡ ∂ ⎛ ∂ ⎛ ∂ ⎛ 11 ∂ϕ ⎞ 22 ∂ϕ ⎞ 33 ∂ϕ ⎞ ⎢ ⎜ g λg ⎟+ ⎜ g λg ⎟+ ⎜ g λg ⎟⎥ ∂ξ ⎠ ∂η ⎝ ∂η ⎠ ∂ζ ⎝ ∂ζ ⎠ ⎥⎦ g ⎣⎢ ∂ξ ⎝
1 ∂ ⎛ ∂ϕ ⎞ 1 ∂ ⎛ 1 ∂ϕ ⎞ ∂ ⎛ ∂ϕ ⎞ ⎜ ξλ ⎟+ ⎜λ ⎟+ ⎜λ ⎟. ξ ∂ξ ⎝ ∂ξ ⎠ ξ ∂η ⎝ ξ ∂η ⎠ ∂ζ ⎝ ∂ζ ⎠
(250)
This is the well-known form in cylindrical coordinates – compare for instance with Bird, Stewart and Lightfood (1960), p. 739B
700
Appendix 2 Basics of the coordinate transformation theory
Spherical coordinates: x = ξ sin η cos ζ ,
ξ = + x2 + y 2 + z 2 ,
y = ξ sin η sin ζ ,
x = arctan
z = ξ cos η ,
ζ = arctan ( y x ) .
x2 + y 2 z
(251,252) ,
(253,254) (255,256)
Here ξ is the radial coordinate, η is the azimuthal coordinate and ζ is the polar coordinate. An equidistant transformation in the computational space is defined by ⎡ ξ ⎤ ⎛ θ − θ min ⎞ ⎛ ϕ max − ϕ min ⎞ x (ξi ,η j ) = ⎢ rmin + ( rmax − rmin ) i ⎥ sin ⎜ max η j ⎟ cos ⎜ ζ j ⎟, imax ⎦ ⎝ jmax kmax ⎣ ⎠ ⎝ ⎠ (257) ⎡ ξ ⎤ ⎛ θ − θ min ⎞ ⎛ ϕ max − ϕ min ⎞ y (ξi ,η j ) = ⎢ rmin + ( rmax − rmin ) i ⎥ sin ⎜ max η j ⎟ sin ⎜ ζ j ⎟, imax ⎦ ⎝ jmax kmax ⎣ ⎠ ⎝ ⎠ (258) ⎡ ξ ⎤ ⎛ θ − θ min ⎞ z (ξi ,η j ) = ⎢ rmin + ( rmax − rmin ) i ⎥ cos ⎜ max ηj ⎟, i jmax ⎣ max ⎦ ⎝ ⎠
(259) where
ξi = 0,1, 2,..., imax , η j = 0,1, 2,..., jmax , ζ k = 0,1, 2,..., kmax . The inverse metrics of the coordinate transformation are then ⎛ ∂x ⎜ ⎜ ∂ξ ⎜ ∂y ⎜ ⎜ ∂ξ ⎜ ∂z ⎜⎜ ⎝ ∂ξ
∂x ∂η ∂y ∂η ∂z ∂η
∂x ∂ζ ∂y ∂ζ ∂z ∂ζ
⎞ ⎟ ⎟ ⎛ sin η cos ζ ⎟ ⎜ ⎟ = ⎜ sin η sin ζ ⎟ ⎜ cos η ⎟ ⎝ ⎟⎟ ⎠
ξ cos η cos ζ ξ cos η sin ζ −ξ sin η
−ξ sin η sin ζ ξ sin η cos ζ 0
⎞ ⎟ ⎟. ⎟ ⎠
(260)
The Jacobian is then ⎛ ∂y ∂z ∂y ∂z ⎞ ∂x ⎛ ∂y ∂z ∂y ∂z ⎞ ∂x ⎛ ∂y ∂z ∂y ∂z ⎞ − − − ⎜ ⎟− ⎜ ⎟+ ⎜ ⎟ ⎝ ∂η ∂ζ ∂ζ ∂η ⎠ ∂η ⎝ ∂ξ ∂ζ ∂ζ ∂ξ ⎠ ∂ζ ⎝ ∂ξ ∂η ∂η ∂ξ ⎠ = ξ 2 sin η . (261) g =
∂x ∂ξ
Appendix 2 Basics of the coordinate transformation theory
The metrics of the coordinate transformation are ⎛ ∂ξ ∂ξ ∂ξ ⎞ ⎛ ∂y ∂z ∂y ∂z ∂x ∂z ∂x ∂z − − ⎜ ⎟ ⎜ ⎜ ∂x ∂y ∂z ⎟ ⎜ ∂η ∂ζ ∂ζ ∂η ∂ζ ∂η ∂η ∂ζ ⎜ ∂η ∂η ∂η ⎟ 1 ⎜ ∂y ∂z ∂y ∂z ∂x ∂z ∂x ∂z − − ⎜ ⎟= ⎜ g ⎜ ∂ζ ∂ξ ∂ξ ∂ζ ∂ξ ∂ζ ∂ζ ∂ξ ⎜ ∂x ∂y ∂z ⎟ ⎜ ∂ζ ∂ζ ∂ζ ⎟ ⎜ ∂y ∂z ∂y ∂z ∂x ∂z ∂x ∂z − − ⎜⎜ ⎟⎟ ⎜⎜ ⎝ ∂x ∂y ∂z ⎠ ⎝ ∂ξ ∂η ∂η ∂ξ ∂η ∂ξ ∂ξ ∂η
⎛ ⎜ ⎜ sin η cos ζ ⎜1 = ⎜ cos ζ cosη ⎜ξ ⎜ 1 sin ζ ⎜⎜ − ξ sin η ⎝
sin η sin ζ 1
ξ
sin ζ cosη 1 cos ζ ξ sin η
∂x ∂η ∂x ∂ζ ∂x ∂ξ
⎞ ⎟ cosη ⎟ ⎟ 1 − sin η ⎟ . ξ ⎟ ⎟ 0 ⎟⎟ ⎠
701
∂y ∂x ∂y ⎞ − ⎟ ∂ζ ∂ζ ∂η ⎟ ∂y ∂x ∂y ⎟ − ⎟ ∂ξ ∂ξ ∂ζ ⎟ ∂y ∂x ∂y ⎟ − ⎟ ∂η ∂η ∂ξ ⎟⎠
(262)
The components of the three covariant base vectors of the curvilinear coordinate system are the columns of the matrix of the inverse metrics. They are given below together with the corresponding covariant unit vectors. ∂x ∂y ∂z i+ j+ k = sin η cos ζ i + sin η sin ζ j + cos η k , ∂ξ ∂ξ ∂ξ e1 = sin η cos ζ i + sin η sin ζ j + cos ηk , a1 =
a1 = 1, (263-265)
∂x ∂y ∂z i+ j+ k = ξ cos η cos ζ i + ξ cos η sin ζ j − ξ sin ηk , a 2 = ξ , ∂η ∂η ∂η e2 = cos η cos ζ i + cos η sin ζ j − sin ηk , (266-268) a2 =
∂x ∂y ∂z i+ j+ k = −ξ sin η sin ζ i + ξ sin η cos ζ j , ∂ζ ∂ζ ∂ζ e3 = − sin ζ i + cos ζ j. a3 =
a3 = ξ sin η , (269-271)
The components of the three contravariant base vectors of the curvilinear coordinate system are the rows of the matrix of the metrics. They are given below together with the corresponding contravariant unit vectors. a1 =
∂ξ ∂ξ ∂ξ i+ j+ k = sin η cos ζ i + sin η sin ζ j + cosη k , ∂x ∂y ∂z
e1 = sin η cos ζ i + sin η sin ζ j + cos ηk , a2 =
a1 = 1,
(272-274)
∂η ∂η ∂η 1 1 1 1 i+ j+ k = cos ζ cosη i + sin ζ cosη j − sin η k , a 2 = , ∂x ∂y ∂z ξ ξ ξ ξ
e2 = cos ζ cos η i + sin ζ cos η j − sin η k ,
(275-277)
702
Appendix 2 Basics of the coordinate transformation theory
a3 =
∂ζ ∂ζ ∂ζ 1 sin ζ 1 cos ζ i+ j+ k=− i+ j, ∂x ∂y ∂z ξ sin η ξ sin η
a3 =
e3 = − sin ζ i + cos ζ j.
1 , ξ sin η
(278-280)
We see that a) the two types of base vectors are parallel and that b) three base vectors of each type are mutually perpendicular which means that the cylindrical coordinate system is orthogonal. ∇⋅F = =
1 ⎡ ∂ ⎢ g ⎣ ∂ξ
(
)
g a1 e1 ⋅ F +
∂ ∂η
(
)
g a 2 e2 ⋅ F +
∂ ∂ζ
⎡ ∂ ⎤ 1 ∂ ∂ ξ 2 sin η Fξ + ξ sin η Fη + ξ Fζ ⎥ . ⎢ ξ sin η ⎣ ∂ξ ∂η ∂ζ ⎦
(
2
)
(
)
(
)
(
⎤ g a1 e3 ⋅ F ⎥ ⎦
)
(281)
This is the expected form for spherical coordinates – compare for instance with Bird, Stewart and Lightfood (1960), p. 739C. The covariant metric tensor is ⎛ a1 ⋅ a1 a1 ⋅ a 2 ⎜ gij = ⎜ a 2 ⋅ a1 a 2 ⋅ a 2 ⎜ a ⋅a a ⋅a 3 2 ⎝ 3 1
a1 ⋅ a3 ⎞ ⎛ 1 0 0 ⎞ ⎟ ⎜ ⎟ 2 a 2 ⋅ a3 ⎟ = ⎜ 0 ξ 0 ⎟, 2 2 ⎟ ⎜ a3 ⋅ a3 ⎠ ⎝ 0 0 ξ sin η ⎟⎠
(282)
which is a symmetric tensor. The angles between the covariant unit vectors are
θ12 = arccos ⎜
⎛
g12
⎞ π ⎟= , ⎟ 2 ⎠
(283)
θ13 = arccos ⎜
⎛
g13
⎞ π ⎟= , ⎟ 2 ⎠
(284)
⎛
g 23
⎞ π ⎟= . ⎜ g 22 g 33 ⎟ 2 ⎝ ⎠
(285)
⎜ g11 g 22 ⎝
⎜ g11 g33 ⎝
θ 23 = arccos ⎜
The magnitude of the increment of the surface area is
ξ = const ,
2 dS 1 = g 22 g 33 − g 23 dη d ζ = ξ 2 sin η dη d ζ ,
(286)
η = const ,
dS 2 = g11 g33 − g132 d ξ dζ = ξ sin η d ξ d ζ ,
(287)
ζ = const ,
dS 3 = g11 g 22 − g122 dξ dη = ξ dξ dη .
(288)
Appendix 2 Basics of the coordinate transformation theory
703
The volume increment is then dV = g dξ dη d ζ = ξ 2 sin η dξ dη dζ .
(289)
The elements of the contravariant metric tensor are ⎛ a1 ⋅ a1 a1 ⋅ a 2 ⎜ g ij = ⎜ a 2 ⋅ a1 a 2 ⋅ a 2 ⎜ a3 ⋅ a1 a3 ⋅ a 2 ⎝ 2 ⎛ g 22 g33 − g 23 1⎜ = ⎜ g 21 g 33 − g 23 g 31 g⎜ ⎝ g 21 g 32 − g 22 g 31
a1 ⋅ a 3 ⎞ ⎟ a 2 ⋅ a3 ⎟ a3 ⋅ a3 ⎟⎠
g 21 g33 − g 23 g31 g11 g 33 − g132 g11 g 32 − g12 g 31
g 21 g32 − g 22 g 31 ⎞ ⎟ g11 g 32 − g12 g 31 ⎟ g11 g 22 − g122 ⎟⎠
⎛ ⎜ ⎜1 0 ⎜ 1 = ⎜0 2 ξ ⎜ ⎜ ⎜⎜ 0 0 ⎝
⎞ ⎟ 0 ⎟ ⎟ 0 ⎟. ⎟ ⎟ 1 ⎟ 2 2 ξ sin η ⎟⎠ (290)
With this result we can compute the Diffusion Laplacian of a scalar for the orthogonal system ∇ ⋅ ( λ∇ϕ ) = =
⎤ 1 ⎡ ∂ ⎛ ∂ ⎛ ∂ ⎛ 11 ∂ϕ ⎞ 22 ∂ϕ ⎞ 33 ∂ϕ ⎞ ⎢ ⎜ g λg ⎟+ ⎜ g λg ⎟+ ⎜ g λg ⎟⎥ ∂ξ ⎠ ∂η ⎝ ∂η ⎠ ∂ζ ⎝ ∂ζ ⎠ ⎥⎦ g ⎢⎣ ∂ξ ⎝ 1 ∂ ⎛ 2 ∂ϕ ⎞ 1 ∂ ⎛ ∂ϕ ⎞ 1 ∂ ⎛ ∂ϕ ⎞ ⎜ξ λ ⎟+ ⎜ sin ηλ ⎟+ ⎜λ ⎟. ∂ξ ⎠ ξ 2 sin η ∂η ⎝ ∂η ⎠ ξ 2 sin 2 η ∂ζ ⎝ ∂ζ ⎠ (291)
ξ 2 ∂ξ ⎝
This is the well-known form in spherical coordinates – compare for instance with Bird, Stewart and Lightfood (1960), p. 739C. Elliptic grid generation systems: The most popular grid generating system makes use of the Laplace equation. The solution of the Laplace equations (harmonic functions) obeys minimum and maximum values only at the boundary (maximum principle) and is smooth (derivatives of all orders exists). One dimensional stretching along the three axes correspondingly, delivers also a solution which has the similar properties as the solution of the Laplace equation. The later is proven to be equivalent to solving the Poison-like equations, Thompson, Warsi and Mastin (1985),
ξ xx + ξ yy + ξ zz = P (ξ ,η , ζ ) ,
(292)
η xx + η yy + η zz = Q (ξ ,η , ζ ) ,
(293)
704
Appendix 2 Basics of the coordinate transformation theory
ζ xx + ζ yy + ζ zz = R (ξ ,η , ζ ) ,
(294)
with boundary conditions prescribed at the surface containing the domain (Dirichlet boundary value problem). The control functions P, Q and R define the concentration of the grid lines to a prescribed lines or points. Interchanging the dependent (ξ ,η , ζ ) and the independent ( x, y , z ) variables results in ⎛ ∂ 2r ∂2r ∂ 2r ∂2r ∂ 2r ∂ 2r ⎞ + g 22 + g 33 + 2 ⎜ g 12 + g 13 + g 23 ⎟ 2 2 2 ∂ξ ∂η ∂ζ ∂ξ∂η ∂ξ∂ζ ∂η∂ζ ⎠ ⎝ ∂r ∂r ∂r +P +Q +R = 0, ∂ξ ∂η ∂ζ g 11
or rewritten in components form 2 2 2 2 ⎛ 12 ∂ 2 x ∂2 x 22 ∂ x 33 ∂ x 13 ∂ x 23 ∂ x ⎞ + g + g + 2 g + g + g ⎜ ⎟ ∂ξ 2 ∂η 2 ∂ζ 2 ∂ξ∂η ∂ξ∂ζ ∂η∂ζ ⎠ ⎝ ∂x ∂x ∂x +P +Q +R = 0, (295) ∂ξ ∂η ∂ζ
g 11
⎛ ∂2 y ∂2 y ∂2 y ∂2 y ∂2 y ∂2 y ⎞ + g 22 + g 33 + 2 ⎜ g 12 + g 13 + g 23 ⎟ 2 2 2 ∂ξ ∂η ∂ζ ∂ξ∂η ∂ξ∂ζ ∂η∂ζ ⎠ ⎝ ∂y ∂y ∂y +P +Q +R = 0, (296) ∂ξ ∂η ∂ζ g 11
⎛ ∂2 z ∂2 z ∂2 z ∂2 z ∂2 z ∂2 z ⎞ + g 22 + g 33 + 2 ⎜ g 12 + g 13 + g 23 ⎟ 2 2 2 ∂ξ ∂η ∂ζ ∂ξ∂η ∂ξ∂ζ ∂η∂ζ ⎠ ⎝ ∂z ∂z ∂z +P +Q +R = 0, (297) ∂ξ ∂η ∂ζ g 11
where 2 ⎛ g 22 g33 − g 23 1⎜ g = ⎜ g 21 g 33 − g 23 g 31 g⎜ ⎝ g 21 g32 − g 22 g31
ij
g 21 g33 − g 23 g31 g11 g 33 − g132 g11 g32 − g12 g31
g 21 g 32 − g 22 g 31 ⎞ ⎟ g11 g 32 − g12 g 31 ⎟ g11 g 22 − g122 ⎟⎠
(298)
are the elements of the contravariant metric tensor which is symmetric, Thompson, Warsi and Mastin (1985). The elements of the contravariant metric tensor are computed as a function of the elements of the covariant metric tensor. The covariant metric tensor is defined as
Appendix 2 Basics of the coordinate transformation theory
⎛ a1 ⋅ a1 a1 ⋅ a 2 ⎜ g ij = ⎜ a 2 ⋅ a1 a 2 ⋅ a 2 ⎜ a ⋅a a ⋅a 3 2 ⎝ 3 1
705
a1 ⋅ a3 ⎞ ⎟ a 2 ⋅ a3 ⎟ a3 ⋅ a3 ⎟⎠
⎛ ⎛ ∂x ⎞ 2 ⎛ ∂y ⎞ 2 ⎛ ∂z ⎞ 2 ⎜⎜ ⎟ +⎜ ⎟ +⎜ ⎟ ⎜ ⎝ ∂ξ ⎠ ⎝ ∂ξ ⎠ ⎝ ∂ξ ⎠ ⎜ ⎜ ∂x ∂x ∂y ∂y ∂z ∂z =⎜ + + ⎜ ∂ξ ∂η ∂ξ ∂η ∂ξ ∂η ⎜ ⎜ ∂x ∂x + ∂y ∂y + ∂z ∂z ⎜ ∂ξ ∂ζ ∂ξ ∂ζ ∂ξ ∂ζ ⎝
∂x ∂x ∂y ∂y ∂z ∂z + + ∂ξ ∂η ∂ξ ∂η ∂ξ ∂η 2
2
⎛ ∂x ⎞ ⎛ ∂y ⎞ ⎛ ∂z ⎞ ⎜ ⎟ +⎜ ⎟ +⎜ ⎟ ⎝ ∂η ⎠ ⎝ ∂η ⎠ ⎝ ∂η ⎠
2
∂x ∂x ∂y ∂y ∂z ∂z + + ∂η ∂ζ ∂η ∂ζ ∂η ∂ζ
∂x ∂x ∂y ∂y ∂z ∂z ⎞ ⎟ + + ∂ξ ∂ζ ∂ξ ∂ζ ∂ξ ∂ζ ⎟ ⎟ ∂x ∂x ∂y ∂y ∂z ∂z ⎟ + + ⎟, ∂η ∂ζ ∂η ∂ζ ∂η ∂ζ ⎟ 2 2 2 ⎟ ⎛ ∂x ⎞ ⎛ ∂y ⎞ ⎛ ∂z ⎞ ⎟ ⎜ ⎟ +⎜ ⎟ +⎜ ⎟ ⎝ ∂ζ ⎠ ⎝ ∂ζ ⎠ ⎝ ∂ζ ⎠ ⎟⎠
(299) where a1 =
∂x ∂y ∂z i+ j+ k, ∂ξ ∂ξ ∂ξ
(300)
a2 =
∂x ∂y ∂z i+ j+ k, ∂η ∂η ∂η
(301)
a3 =
∂x ∂y ∂z i+ j+ k. ∂ζ ∂ζ ∂ζ
(302)
In fact all coefficients are functions of the elements in the Jacobian matrix of the coordinate transformation x = f (ξ ,η , ζ ) , y = g (ξ ,η , ζ ) , z = h (ξ ,η , ζ ) , ⎛ ∂x ⎜ ⎜ ∂ξ ⎜ ∂y J (ξ , η , ζ ) = ⎜ ⎜ ∂ξ ⎜ ∂z ⎜⎜ ⎝ ∂ξ
∂x ∂η ∂y ∂η ∂z ∂η
∂x ∂ζ ∂y ∂ζ ∂z ∂ζ
⎞ ⎟ ⎟ ⎟ ⎟. ⎟ ⎟ ⎟⎟ ⎠
(303)
The elements of the Jacobian matrix are called inverse metrics of the coordinate transformation. The Jacobian determinant or Jacobian of the coordinate transformation is g = J (ξ ,η , ζ ) =
∂x ⎛ ∂y ∂z ∂y ∂z ⎞ ∂x ⎛ ∂y ∂z ∂y ∂z ⎞ ∂x ⎛ ∂y ∂z ∂y ∂z ⎞ − − − ⎜ ⎟− ⎜ ⎟+ ⎜ ⎟. ∂ξ ⎝ ∂η ∂ζ ∂ζ ∂η ⎠ ∂η ⎝ ∂ξ ∂ζ ∂ζ ∂ξ ⎠ ∂ζ ⎝ ∂ξ ∂η ∂η ∂ξ ⎠ (304)
706
Appendix 2 Basics of the coordinate transformation theory
For two dimensional case for instance for not existing j-direction set ∂x j ξ j = 1 , ∂xi ξ j ≠ i = 0 . A simple prescription for computation of gg ij is 3
gg ij = α ij = ∑ Ami Amj ,
(298a)
m =1
Miki and Takagi (1984), where Ami is the mith cofactor of the Jacobian matrix of the coordinate transformation, Eq. (303). Ami = ( Subdeterminant mi )( −1)
m +i
.
Boundary conditions: The surface containing the domain of interest is divided into 6 non-overlapping patches, Γi (i = 1,2,3,4,5,6), building 3 pairs perpendicular to each transformed directions. The boundary conditions, that is, the coordinates at the 6 patches are prescribed in advance as follows ⎛ x ⎞ ⎛ f i (ξ i , η , ζ ) ⎞ ⎟ ⎜ ⎟ ⎜ ⎜ y ⎟ = ⎜ g i (ξ i , η , ζ ) ⎟ , ⎜ z ⎟ ⎜ h (ξ , η , ζ ) ⎟ ⎝ ⎠ ⎝ i i ⎠
(ξi ,η , ζ ) ∈ Γi , ( i = 1, 2 ) ,
(305)
⎛ x ⎞ ⎛ f i (ξ , ηi , ζ ) ⎞ ⎟ ⎜ ⎟ ⎜ ⎜ y ⎟ = ⎜ g i (ξ , ηi , ζ ) ⎟ , ⎜ z ⎟ ⎜ h (ξ , η , ζ ) ⎟ i ⎝ ⎠ ⎝ i ⎠
(ξ ,ηi , ζ ) ∈ Γi , ( i = 3, 4 ) ,
(306)
⎛ x ⎞ ⎛ f i (ξ , η , ζ i ) ⎞ ⎟ ⎜ ⎟ ⎜ ⎜ y ⎟ = ⎜ g i (ξ , η , ζ i ) ⎟ , ⎜ z ⎟ ⎜ h (ξ , η , ζ ) ⎟ i ⎠ ⎝ ⎠ ⎝ i
(ξ ,η , ζ i ) ∈ Γi , ( i = 5, 6 ) .
(307)
Line spacing control functions: The line spacing control functions n
P (ξ ,η , ζ ) = −∑ al1 sign (ξ − ξl ) exp ( −bl1Tl ) ,
(308)
l =1 n
Q (ξ ,η , ζ ) = −∑ al 2 sign (η − ηl ) exp ( −bl 2Tl ) ,
(309)
l =1 n
R (ξ ,η , ζ ) = −∑ al 3 sign (ζ − ζ l ) exp ( −bl 3Tl ),
(310)
l =1
where T = cl1 (ξ − ζ l ) + cl 2 (η − ηl ) + cl 3 (ζ − ζ l ) , 2
2
2
(311)
Appendix 2 Basics of the coordinate transformation theory
707
are recommended by Miki and Takagi (1984). The sets of the coefficients alm and clm attracts the ξ , η , and/or ζ = constant planes to the specified plane, coordinate line, or grid point. The range of the attraction effect is determined by the decay factor blm . The effect of the coefficient is demonstrated below: Attracted planes
Planes, line, or point attracting neighboring planes
al1 ≠ 0 for ξ =const plane
al1 ≠ 0 , cl1 ≠ 0 , al1 = 0 , cl1 = 0 , al1 = 0 , cl1 = 0 ,
al 2 ≠ 0 for η =const plane al 3 ≠ 0 for ζ =const plane
al 2 = 0 , al 3 = 0 for ξ = ξl plane cl 2 = 0 , cl 3 = 0 al 2 ≠ 0 , al 3 = 0 for η = ηl plane cl 2 ≠ 0 , cl 3 = 0 al 2 = 0 , al 3 ≠ 0 for ζ = ζ l plane cl 2 = 0 , cl 3 ≠ 0
For (η , ζ ) = (ηl , ζ l ) line: al1 = 0 , al 2 ≠ 0 , al 3 ≠ 0 cl1 = 0 , cl 2 ≠ 0 , cl 3 ≠ 0
For (ξ , ζ ) = (ξl , ζ l ) line: al1 = 0 , al 2 = 0 , al 3 ≠ 0 cl1 = 0 , cl 2 = 0 , cl 3 ≠ 0
For (ξ , η ) = (ξ l ,ηl ) line: al1 ≠ 0 , al 2 ≠ 0 , al 3 = 0 cl1 ≠ 0 , cl 2 ≠ 0 , cl 3 = 0
For (ξ , η , ζ ) = (ξl ,ηl , ζ l ) point: al1 ≠ 0 , al 2 ≠ 0 , al 3 ≠ 0 cl1 ≠ 0 , cl 2 ≠ 0 , cl 3 ≠ 0 Simplification for 2D plane: For grid generation in 2D plane the grid generating system (ξ ,η ) simplifies to g 11
∂2 x ∂2 x ∂2 x ∂x ∂x + g 22 + 2 g 12 +P +Q = 0, 2 2 ∂ξ ∂η ∂ξ∂η ∂ξ ∂η
(312)
g 11
∂2 y ∂2 y ∂2 y ∂y ∂y + g 22 + 2 g 12 +P +Q = 0, 2 2 ∂ξ ∂η ∂ξ∂η ∂ξ ∂η
(313)
708
Appendix 2 Basics of the coordinate transformation theory
where
g ij =
1 ⎛ g 22 ⎜ g ⎝ g 21
⎛ ⎛ ∂x ⎞ 2 ⎛ ∂y ⎞ 2 ⎜⎜ ⎟ +⎜ ⎟ g 21 ⎞ 1 ⎜ ⎝ ∂η ⎠ ⎝ ∂η ⎠ ⎟= ⎜ g11 ⎠ g ⎜ ∂x ∂x ∂y ∂y ⎜ ∂ξ ∂η + ∂ξ ∂η ⎝
∂x ∂x ∂y ∂y ⎞ ⎟ + ∂ξ ∂η ∂ξ ∂η ⎟ , 2 2 ⎟ ⎛ ∂x ⎞ ⎛ ∂y ⎞ ⎟ ⎜ ⎟ +⎜ ⎟ ⎟ ⎝ ∂ξ ⎠ ⎝ ∂ξ ⎠ ⎠
(314)
and g =
∂x ∂y ∂x ∂y − . ∂ξ ∂η ∂η ∂ξ
(315)
Numerical solution: Central difference numerical scheme is applied for computation of the partial derivatives assuming Δξ = Δη = Δζ = 1 : ∂2 x 1 = ( xi +1 − 2 x + xi −1 ), ∂ξ 2 4
∂x 1 = ( xi +1 − xi −1 ) , ∂ξ 2
(316, 317)
∂2 x 1 = x j +1 − 2 x + x j −1 , ∂η 2 4
∂x 1 = x j +1 − x j −1 , ∂η 2
(318, 319)
∂2 x 1 = ( xk +1 − 2 x + xk −1 ) , ∂ζ 2 4
∂x 1 = ( xk +1 − xk −1 ) , ∂ζ 2
(320, 321)
∂2 y 1 = ( yi +1 − 2 y + yi −1 ), ∂ξ 2 4
∂x 1 = ( yi +1 − yi −1 ), ∂ξ 2
(322, 323)
∂2 y 1 = y j +1 − 2 y + y j −1 , ∂η 2 4
∂x 1 = y j +1 − y j −1 , ∂η 2
(324, 325)
∂2 y 1 = ( yk +1 − 2 y + yk −1 ) , ∂ζ 2 4
∂y 1 = ( yk +1 − yk −1 ) , ∂ζ 2
(326, 327)
(
(
)
)
(
)
(
)
∂2 z 1 = ( zi +1 − 2 z + zi −1 ) , ∂ξ 2 4
∂z 1 = ( zi +1 − zi −1 ) , ∂ξ 2
(326, 327)
∂2 z 1 = z j +1 − 2 z + z j −1 , ∂η 2 4
∂z 1 = z j +1 − z j −1 , ∂η 2
(328, 329)
∂2 z 1 = ( zk +1 − 2 z + zk −1 ) , ∂ζ 2 4
∂z 1 = ( zk +1 − zk −1 ), ∂ζ 2
(330, 331)
(
)
(
)
Appendix 2 Basics of the coordinate transformation theory
709
∂2 x ∂ ∂x 1 = = xi +1, j +1 − xi +1, j −1 − xi −1, j +1 + xi −1, j −1 , ∂ξ∂η ∂ξ ∂η 4
(332, 333)
∂2 x ∂ ∂x 1 = = ( xi +1, k +1 − xi +1, k −1 − xi −1, k +1 + xi −1, k −1 ) , ∂ξ∂ζ ∂ξ ∂ζ 4
(334, 335)
(
)
∂2 x ∂ ∂x 1 = = x j +1, k +1 − x j +1, k −1 − x j −1, k +1 + x j −1, k −1 , ∂η∂ζ ∂η ∂ζ 4
)
(336, 337)
∂2 y ∂ ∂y 1 = = yi +1, j +1 − yi +1, j −1 − yi −1, j +1 + yi −1, j −1 , ∂ξ∂η ∂ξ ∂η 4
(339, 340)
∂2 y ∂ ∂y 1 = = ( yi +1, k +1 − yi +1, k −1 − yi −1, k +1 + yi −1, k −1 ), ∂ξ∂ζ ∂ξ ∂ζ 4
(341, 342)
(
(
)
∂2 y ∂ ∂y 1 = = y j +1, k +1 − y j +1, k −1 − y j −1, k +1 + y j −1, k −1 , ∂η∂ζ ∂η ∂ζ 4
(343, 344)
∂2 z ∂ ∂z 1 = = zi +1, j +1 − zi +1, j −1 − zi −1, j +1 + zi −1, j −1 , ∂ξ∂η ∂ξ ∂η 4
(345, 346)
∂2 z ∂ ∂z 1 = = ( zi +1, k +1 − zi +1, k −1 − zi −1, k +1 + zi −1, k −1 ) , ∂ξ∂ζ ∂ξ ∂ζ 4
(347, 348)
∂2 z ∂ ∂z 1 = = z j +1, k +1 − z j +1, k −1 − z j −1, k +1 + z j −1, k −1 . ∂η∂ζ ∂η ∂ζ 4
(349, 350)
(
(
(
)
)
)
Replacing in Eq. (312) results in ck −1 xk −1 + c j −1 x j −1 + ci −1 xi −1 + cx + ci +1 xi +1 + c j +1 x j +1 + ck +1 xk +1 = d1 , ck −1 yk −1 + c j −1 y j −1 + ci −1 yi −1 + cy + ci +1 yi +1 + c j +1 y j +1 + ck +1 yk +1 = d 2 , ck −1 zk −1 + c j −1 z j −1 + ci −1 zi −1 + cz + ci +1 zi +1 + c j +1 z j +1 + ck +1 zk +1 = d3 ,
where ck +1 = g 33 2 + R , c j +1 = g 22 2 + Q , ci +1 = g11 2 + P , ck −1 = g 33 2 − R , c j −1 = g 22 2 − Q , ci −1 = g 11 2 − P , c = − ( g 11 + g 22 + g 33 ) = − ( ck −1 + c j −1 + ci −1 + ci +1 + c j +1 + ck +1 ) , d1 = − g 12 ( xi +1, j +1 − xi +1, j −1 − xi −1, j +1 + xi −1, j −1 )
710
Appendix 2 Basics of the coordinate transformation theory
− g 13 ( xi +1, k +1 − xi +1, k −1 − xi −1, k +1 + xi −1, k −1 ) − g 22 ( x j +1, k +1 − x j +1, k −1 − x j −1, k +1 + x j −1, k −1 ) , d 2 = − g 12 ( yi +1, j +1 − yi +1, j −1 − yi −1, j +1 + yi −1, j −1 ) − g 13 ( yi +1, k +1 − yi +1, k −1 − yi −1, k +1 + yi −1, k −1 ) − g 22 ( y j +1, k +1 − y j +1, k −1 − y j −1, k +1 + y j −1, k −1 ) , d 3 = − g 12 ( zi +1, j +1 − zi +1, j −1 − zi −1, j +1 + zi −1, j −1 ) − g13 ( zi +1, k +1 − zi +1, k −1 − zi −1, k +1 + zi −1, k −1 ) − g 22 ( z j +1, k +1 − z j +1, k −1 − z j −1, k +1 + z j −1, k −1 ) ,
The resulting system is solved using SOR methods. Note that the coefficient matrix is the same for all directions. Therefore once the matrix is inverted it can be used for the three right hand sites which safes computer time. Do not forget to treat properly the boundary conditions by putting into the right hand sites the known boundary terms. Orthogonal systems: Numerical orthogonality is reached by setting the offdiagonal elements of the contravariant metric tensor equal to zero. This result in ∂ ∂ξ
⎛ ∂x ⎜⎜ ⎝ ∂ξ
g 22 g 33 g11
⎞ ∂ ⎛ ∂x ⎟⎟ + ⎜⎜ ⎠ ∂η ⎝ ∂η
g11 g 33 g 22
⎞ ∂ ⎟⎟ + ⎠ ∂ζ
⎛ ∂x ⎜⎜ ⎝ ∂ζ
g11 g 22 g 33
⎞ ⎟⎟ = 0 , ⎠
(351)
∂ ∂ξ
⎛ ∂y ⎜⎜ ⎝ ∂ξ
g 22 g 33 g11
⎞ ∂ ⎛ ∂y ⎟⎟ + ⎜⎜ ⎠ ∂η ⎝ ∂η
g11 g 33 g 22
⎞ ∂ ⎟⎟ + ⎠ ∂ζ
⎛ ∂y ⎜⎜ ⎝ ∂ζ
g11 g 22 g 33
⎞ ⎟⎟ = 0 , ⎠
(352)
∂ ∂ξ
⎛ ∂z ⎜⎜ ⎝ ∂ξ
g 22 g 33 g11
⎞ ∂ ⎛ ∂z ⎟⎟ + ⎜⎜ ⎠ ∂η ⎝ ∂η
g11 g 33 g 22
⎞ ∂ ⎟⎟ + ⎠ ∂ζ
⎛ ∂z ⎜⎜ ⎝ ∂ζ
g11 g 22 g 33
⎞ ⎟⎟ = 0 . ⎠
(353)
see Thompson and Warsi (1982). If, in addition to orthogonality, the condition g11 = g 22 = g33 = const is imposed, then the above three equations reduce to Laplace equations. For 2D systems the equations set (351-353) reduces to ∂ ⎛ ∂x ⎞ ∂ ⎛ 1 ∂x ⎞ ⎜ ⎟ = 0, ⎜ fr ⎟+ ∂ξ ⎝ ∂ξ ⎠ ∂η ⎝ f r ∂η ⎠
(354)
∂ ⎛ ∂y ⎞ ∂ ⎛ 1 ∂y ⎞ ⎜ ⎟ = 0, ⎜ fr ⎟+ ∂ξ ⎝ ∂ξ ⎠ ∂η ⎝ f r ∂η ⎠
(355)
where f r = g 22 g11 > 0
(356)
Appendix 2 Basics of the coordinate transformation theory
711
is called “distortion function”. It specifies the ratio of the sides of small rectangle in the x, y plane which is mapped onto a rectangle in the ξ , η plane. A reasonable upper limit for f r is 10. The condition of orthogonality also leads to ∂x ∂y ∂y ∂x = − fr , = fr . ∂η ∂ξ ∂η ∂ξ
(357, 358)
This conditions have to be satisfied also at the boundary. For f r ( ξ , η ) = 1,
(359)
Eqs. (357, 358) become the Cauchy-Riemann equations which are used in conformal mapping. Tamamidis and Assanis (1991) recognized that non uniform f r over the domain of interest cane be used for controlling the spacing of the nodes even generating orthogonal grids. The authors successfully examined the properties of ∂2 fr ∂2 fr + = G (ξ , η ) ∂ξ 2 ∂η 2
(360)
for generating a smooth f r ’s over the domain of interests. G (ξ ,η ) typically involve some combination of sinusoidal, cosinusoidal and exponential functions, i.e., G (ξ , η ) = const f1 (ξ ) f 2 (η ).
(361)
An initial approximation can be obtained by setting f r = 1. The authors generated coefficients of the algebraic systems finite difference to Eqs. (354) and (355) having the same sign by using for the first order differences: forward finite differences for xξ when fξ is positive, and backward differences when fξ is negative; backward finite differences for xη when fη is positive, and forward differences when fη is negative. Setting f = ( g11 g 22 g33 )
−1/ 2
= g −1/ 2
(362)
the equations set (351-353) can be rewritten as ∂ ⎛ ∂x ⎞ ∂ ⎛ ∂x ⎞ ∂ ⎛ ∂x ⎞ ⎜ fg 22 g33 ⎟+ ⎜ fg11 g 33 ⎟+ ⎜ fg11 g 22 ⎟ = 0, ∂ξ ⎝ ∂ξ ⎠ ∂η ⎝ ∂η ⎠ ∂ζ ⎝ ∂ζ ⎠
(363)
∂ ⎛ ∂y ⎞ ∂ ⎛ ∂y ⎞ ∂ ⎛ ∂y ⎞ ⎜ fg 22 g33 ⎟+ ⎜ fg11 g 33 ⎟+ ⎜ fg11 g 22 ⎟ = 0, ∂ξ ⎝ ∂ξ ⎠ ∂η ⎝ ∂η ⎠ ∂ζ ⎝ ∂ζ ⎠
(364)
712
Appendix 2 Basics of the coordinate transformation theory
∂ ⎛ ∂z ⎞ ∂ ⎛ ∂z ⎞ ∂ ⎛ ∂z ⎞ ⎜ fg 22 g33 ⎟+ ⎜ fg11 g 33 ⎟+ ⎜ fg11 g 22 ⎟ = 0. ∂ξ ⎝ ∂ξ ⎠ ∂η ⎝ ∂η ⎠ ∂ζ ⎝ ∂ζ ⎠
(365)
Christov (1982) proposed to use f = f ( x, y, z ) for controlling the spacing of the grid by controlling actually the local value of the Jacobian. For 2D, ∂ ⎛ ∂x ⎞ ∂ ⎛ ∂x ⎞ ⎜ fg 22 ⎟+ ⎜ fg11 ⎟ = 0, ∂ξ ⎝ ∂ξ ⎠ ∂η ⎝ ∂η ⎠
(366)
∂ ⎛ ∂y ⎞ ∂ ⎛ ∂y ⎞ ⎜ fg 22 ⎟+ ⎜ fg11 ⎟ = 0, ∂ξ ⎝ ∂ξ ⎠ ∂η ⎝ ∂η ⎠
(367)
Christov proposed to impose a control function of the type f = 1 + z x2 + z y2 ,
(368)
that is simply the slope of a predefined surface z = z ( x, y ) . The stronger the slope of the chosen surface at a given point the more dense the grid around this point. The Sorenson method for enforcing orthogonality at the boundaries: Sorenson (1989) oriented his development towards creating a multiple interconnected blocks. He simplified the P, Q, and R functions to facilitate attractions to the corresponding outer boundary only by specifying attenuation factors so as not to influence the opposite boundary surface e.g. P (ξ ,η , ζ ) = p1 (η , ζ ) exp ⎡⎣ −a (ξ − 1) ⎤⎦ + p2 (η , ζ ) exp ⎡⎣ − a (ξ max − ξ ) ⎤⎦ + p3 (ξ , ζ ) exp ⎡⎣ −a (η − 1) ⎤⎦ + p4 (ξ , ζ ) exp ⎡⎣ − a (ηmax − η )⎤⎦ + p5 (ξ ,η ) exp ⎡⎣ −a (ζ − 1)⎤⎦ + p6 (ξ ,η ) exp ⎡⎣ − a (ζ max − ζ ) ⎤⎦ .
(369)
By appropriate selection of the p, q, r coefficients he enforced the ortogonality at the boundary. As example consider face 3. At this face the derivatives rξ , rζ , rξζ , rξξ , and rζζ are computed from the point values known at the surface once for the
generation process. The derivatives rη are found from the desired clustering and orthogonality on face 3. The desired orthogonality and clustering is specified by the following three relations rξ ⋅ rη = 0 , rζ ⋅ rη = 0 , rη ⋅ rη = S 2 , where S is the height to be imposed on the cell on the boundary or in expanded form xξ xη + yξ yη + zξ zη = 0,
(370)
xζ xη + yζ yη + zζ zη = 0 ,
(371)
xξ2 + yη2 + zζ2 = S 2.
(372)
Appendix 2 Basics of the coordinate transformation theory
713
Solving the first two equations with respect to xη and yη results in
(
) (x y
ζ
− xζ yξ ,
(
) (x y
ζ
− xζ yξ ,
xη = zη − zξ yζ + zζ yξ yη = zη − xξ zζ + xζ zξ
ξ
ξ
)
(373)
)
(374)
or comparing with the cofactors of the Jacobian matrix of the coordinate transformation xη = zη ( − A12 ) ( − A32 ) ,
(375)
yη = zη ( − A22 ) ( − A32 ) .
(376)
Substituting in the third equation and solving with respect to zη we obatin zη =
SA32 ± A + A222 + A322 2 12
,
(377)
,
(378)
.
(379)
and consequently xη =
yη =
SA12 ± A + A222 + A322 2 12
SA22 ± A + A222 + A322 2 12
The positive sign is used for right-handed and the negative for left-handed coordinate systems. Now we have the derivatives rη . These derivatives can be differentiated to obtain the mixed derivatives rηξ , rηζ . All these derivatives are obtained as a function of the information stored at the surface and have to be computed once over the generation process. The only derivatives lacking is
(r )
ηη i ,1, k
=
−7ri ,1, k + 8ri ,2, k − ri ,3, k 2 ( Δη )
2
−
3rη ,i ,1, k
Δη
,
(380)
which results from the Taylor series. Now the governing three equations written for the point (i,1,k) ∂r ∂r ∂r pi ,1, k + qi ,1, k + ri ,1, k ∂ξ ∂η ∂ζ =−
⎛ 1⎡ ∂ 2r ∂2r ∂ 2r ∂ 2r ∂ 2 r ⎞ ⎤ α 22 ∂ 2 r + 2 α + α + α ⎢α11 2 + α 33 ⎜ 12 ⎟⎥ − 13 23 g ⎣⎢ ∂ξ ∂ξ∂η ∂ξ∂ζ ∂η∂ζ ⎠ ⎦⎥ g ∂η 2 ∂ζ 2 ⎝
714
Appendix 2 Basics of the coordinate transformation theory
can be solved with respect to the unknown coefficients p, q and r. In each successive iteration the value of the derivatives ( rηη ) and consequently the values of i ,1, k
p, q and r are updated until convergence is reached. Practical recommendation: Specify first the 6 non-overlapping patches. Divide the peripheral boundary on 4 patches. Generate the grid on each boundary line. Use simple 2D generating system and generate approximate 2D grid at each patch. Improve the 2D grid by using elliptic 2D generating system. Use simple algebraic system to generate initial guess for the 3D system. Generate 3D grids by using 3D elliptic generating system. If other domain has to be attached to the already generated grid specify the patch at which the new domain will be attached. Then use the same procedure to generate the 3D grid by using the already generated grid on the common patch. Plot the grid and control the result. References Anderson JD (1995) Computational fluid dynamics, McGraw-Hill, Inc., New York Bird RB, Stewart WE and Lightfood EN (1960) Transport phenomena, John Wiley & Sons, New York Christov CI (April 1982) Orthogonal coordinate meshes with manageable Jacobian, Proc. of a Symposium on the Numerical Generation of Curvilinear Coordinate Systems and their Use in the Numerical Solution of Partial Differential Equations, Nashville, Tennessee, pp 885-894, Thompson JF ed Miki K and Takagi T (1984) J. Comp. Phys., vol 53 pp 319-330 Peyret R editor (1996) Handbook of computational fluid mechanics, Academic Press, London Sorenson RL (July 1989) The 3DGRAPE book: theory, user’s manual, examples, NASA Technical Memorandum 102224 Tamamidis P and Assanis DN (1991) J. Comp. Physics, vol 94 pp 437-453 Thomas GB Jr, Finney RL and Weir MD (1998) Calculus and analytic geometry, 9th Edition, Addison-Wesley Publishing Company, Reading, Massachusetts Thompson JF and Warsi ZUA (1982) Boundary-fitted coordinate systems for numerical solutions of partial differential equations – a review, J. Comput. Phys., vol 47 pp 1-108 Thompson JF, Warsi ZUA, Mastin CW (1985) Numerical grid generation, North-Holand, New York-Amsterdam-Oxford Vivand H (1974) Conservative forms of gas dynamics equations, Rech. Aerosp., no 1971-1 pp 65-68 Vinokur M (1974) Conservation equations of gas dynamics in curvilinear coordinate systems, J. Comput. Phys., vol 14 pp 105-125
36"Xkuwcn"fgoqpuvtcvkqp"qh"vjg"ogvjqf"
The method presented in this monograph is implemented in the computer code IVA [1,5,6]. The verification of the method is presented in the last Chapter of Volume II. The purpose of this Chapter is to demonstrate the capability of the method in simulating very complex flows – consisting of molten material, gases and water in all their possible combinations. The cases considered here are already documented and discussed in separate publications and the interested reader should go to the original sources for details, especially if he or she is interested in discussions on the physical phenomena and the conclusions for practical applications. What were not given in these publications are the videos corresponding to the discussed cases. We will proceed here with a short description of the case considered, provide the video, and discuss briefly the methods, which are operating within the mathematical simulations. The visualization tool used here is SONJA [2].
3603"Ognv/ycvgt"kpvgtcevkqpu" Next generation nuclear reactors are designed to sustain any catastrophic melt of the reactor core and to keep the accident localized inside a specially designed reactor building called containment. 360303"Ecugu"3"vq"6" In [3] a postulated in-vessel melt-water interaction is analyzed caused by failure of the crust above the core support plate and below the molten poll. The initial conditions are given below: The lower head is assumed to be filled with residual water, 15 t, at saturation conditions up to the lower support plate. The molten pool consisting of 101 t Corium with initial temperature 3100 K - see Fig. 14.1. The release cross sections for the first three cases are effectively 0.3267, 0.5809 and 0.9076 m², and the release time constants are 16.43, 9.24 and 5.91 s respectively. The release cross sections are defined with an axial permeability of 0.31 multiplied by the total failed cross sections. Making use of the symmetry we simulate the process as being two dimensional by using 1424 computational cells. Note that the distribution plate within the lower head is modeled. The 15% reduction of the venting cross section caused by the 8 core barrel supports is taken into account. Case 4 is similar to the previous case. The only difference is that the total lower support plate is assumed to
716
14 Visual demonstration of the method
be permeable with 0.31-permeability. Because cases 1 to 3 did not lead to any melt elevation at all we decided to consider the very hypothetical case 4 to see whether even under these hypothetical conditions the feared slug formation is possible. The most important results of the simulation are described below. 14.1.1.1 Case 1 Figure 14.1 presents the material relocation as a function of space and time. We see that after 0.1 s the melt reaches the water. The fragmentation produces larger melt interfacial area density and steam production driving the water depletion. After 1 s there is almost no water left in the lower head.
Fig. 14.1 IVA5 EPR in-vessel melt water interaction analysis: Case 1, failure cross section 0.3267 m². Material relocation as a function of space and time
The pressures versus time functions at four different positions along the symmetry axis, given in [3], compared with the material relocation maps reveal the cyclic nature of this kind of interaction. The initial melt-water interaction is not strong enough to completely deplete the water. This occurs after the third interaction. In this case the lower head pressure was not large enough to invert the melt relocation in the cross section of the failure. The main conclusion drawn in [3] for this particular case is that, during the process considered, the total melt mass integrated over all cells with water volume fraction above 0.52 (within the water - gas mixture) was at maximum several kilograms which presents no danger for RPV upper head integrity.
14.1 Melt-water interactions
717
Fig. 14.2 IVA5 EPR in-vessel melt water interaction analysis: Case 2, failure cross section 0.5809 m². Material relocation as a function of space and time
14.1.1.2 Case 2 This case is similar to case 1 with the difference that the water depletion from the lower head happens faster, as demonstrated in Fig. 14.2, where the material relocation as a function of time and space is presented. 14.1.1.3 Case 3 The material relocation during the transient is presented on Fig. 14.3. In [3] the pressure as a function of time at the symmetry axis for four different positions: bottom, below and above the flow distribution plate, and below the lower core support plate is also presented. Again, as in cases 1 and 2, the first interaction was not strong enough to invert the melt relocation. The second interaction is started if the melt reaches the flow distribution plate. The flow distribution plate obviously serves as a trigger. This is evident from the fact that the pressure spikes occur here first and then propagate in all directions. The second interaction succeeded in halting the melt relocation.
718
14 Visual demonstration of the method
0.0
0.1
0.2
0.5
0.6
0.7
1.092
1.29
1.75
0.3
0.8
2.0
0.4
1.0
2.984 s
Fig. 14.3 IVA5 EPR in-vessel melt water interaction analysis: Case 3, failure cross section 0.9076 m². a) Material relocation as a function of space and time; b) Melt being below the lower core support plate as a function of time. Without interaction the mass after 3 s must be about 59 t
14.1 Melt-water interactions
719
The expanding steam creates RT instabilities and disintegrates the melt above the pressure source, relaxing the pressure source. Losing kinetic energy due to the action of gravity, the dispersed melt then settles down. The sideward sloshing of the remaining continuous melt is reflected from the walls. Thus, at about 1.0 s the melt starts to flow into the lower head again. The water was depleted from the lower head after the second interaction. Part of the water entered the primary circuit but part was reflected from the upper closure of the down-comer. At about 1.5 s both streams of melt and water penetrate each other and the third, most violent interaction is observed. This produces enough pressure to blow the water completely up from the lower head, but not enough to completely invert the melt stream - it stops it for about 0.3 s. The melt relocation then starts again and continues undisturbed. The main conclusion drawn in [3] for this particular case is that the possible energy releases in this case are below the “no deformation limit” and therefore present no danger for the integrity of the containment. We would like to emphasize the cyclic nature of this kind of in-vessel interaction. The first interaction is not always the strongest one. The next ones may be much stronger in so far as water is still available in the lower head. The water is depleted by cyclic explosions, down to the complete removal from the lower head and the down-comer. The available large venting area makes this possible. 14.1.1.4 Case 4 As expected, case 4 gives the most violent interaction compared to the previous three cases. The first explosion after 0.2 s creates enough pressure to accelerate the melt upwards. The melt reaches the top and starts to drop down. After 0.7 s the second explosion takes place as a result of entrapment. The second interaction produces pressure spikes of up to 25 bar in the region of most intensive interaction. The reason for the violent explosion is the dramatic increase in the interfacial melt area. The result is violent acceleration of the melt upwards again. There is a third cycle of the interaction after 2.2 s. The melt falling down accumulates in the external region of the lower head and intermixes again with the remaining water. The result is a third pressure excursion. Again as in the previous cases we observe cyclic interactions. Almost the total amount of water is depleted outside the reactor pressure vessel (RPV) after the third second. The melt mass inside the RPV in three-phase bubbly flow is not larger than 120 kg. The melt inside the RPV that is in “stoichiometric” three-phase flow is after 0.25 s in its major part of the in dispersed droplet-droplet state. 14.1.1.5 Model elements addressed in cases 1 to 4 1. 2.
All conservation equations for field 1 and field 3 including particle number conservation equations; Geometry description using direction-permeabilities;
720
14 Visual demonstration of the method
Fig. 14.4 IVA5 EPR in-vessel melt water interaction analysis: Case 4, failure cross sectionlower support plate with permeability of 0.31. Material relocation as a function of space. Parameter - time: 0, 0.182, 0.256, 0.502, 0.893, 1.360, 1.515, 1.903, 2.206, 2.279, 2.500, and 2.993
14.1 Melt-water interactions
3. 4. 5. 6.
721
Numerical method resolving short pressure spikes and complex material relocations as presented in [3]; Constitutive relations for mass, momentum, and heat transfer for bubbles, droplets and molten particles; Constitutive relations for radiation interaction between very hot melt and surrounding water-steam-air mixture; Constitutive relations controlling the dynamics fragmentation and coalescence.
14.1.1.6 Available file on CD fig_14.04.gif, fig_14.04.html 360304"Ecugu"7."8"cpf"9 In [4] an ex-vessel melt-water interaction caused by symmetric lower head unzipping within the reactor pit was analyzed. The melt-water interaction outside the reactor pressure vessel (RPV) in a typical pressurized water reactor (PWR) during a severe accident is associated with the particular geometry of each reactor system and the applied accident management strategy. If melt is relocated in the lower head (LH) the nature of the buoyancy driven convection causes the largest heat flux at the circumferential level of the melt surface. The heat flux decreases with decreasing arc distance measured from the lowest point of the vessel. In this situation pressure inside the vessel or the stress caused by the weight of the melt itself causes circumferential failure [4]. It is interesting to know what will happen if water is present at the time of lower head failure caused by melt attack for typical PWR. For this purpose we selected the following two hypothetical cases: Case 5 Lower head unzipping and relocation in cavity partially filled with water. No water is available in the RPV. Case 6 Lower head unzipping and relocation in dry cavity. Water is available in the RPV. In case 5 the melt is relocated into the lower head, the region around the reactor is free of water but there is some residual water in the pit below the reactor. Without external cooling ablation takes place in this plane within 10 to 20 min. The result is unzipping of the lower head. The lower head relocation due to gravity is limited by the bottom of the pit. The water in the pit is displaced upwards. Part of the water comes into contact with the melt exercising large scale relocation and surface instabilities caused by the impact. In case 6 injection of water takes place after the melt relocation into the lower head. The injection does not stop the ablation process and the unzipping occurs just as in case 5. The lower head falls into the dry pit. The accelerated liquids are the melt and the water, the later being in a film boiling above the melt. The large scale melt relocation (sloshing) process starts and stratified melt-water interaction takes place.
722
14 Visual demonstration of the method
14.1.2.1 Initial conditions We specify the following initial conditions for the two cases: Case 5 In this case the lower head with the melt inside it is accelerated due to gravitation only up to 5.3 m/s velocity before the lower head reaches the pit bottom. The lower head contains 104 t oxide melt. The mass of the water in the reactor pit is 53 t. The temperature of the water is equal to the saturation temperature at pressure 1 bar, which is assumed as initial system pressure. Case 6 In this case the lower head with the melt inside it is accelerated due to gravitation only up to 5.3 m/s velocity before the lower head reaches the pit bottom. The lower head contains 104 t melt. The mass of the water inside the RPV is 27 t. The temperature of the water is equal to the saturation temperature at pressure 1 bar, which is assumed as initial system pressure. The volumetric fraction of the steam inside the water is 0.3. 14.1.2.2 Driving instability for Case 5 and 6 At the moment at which the accelerated lower head hits the bottom of the pit the stagnation at the walls create pressures which are reflected to the melt surface. The surface acts as reflector again and so pressure waves interfere, resulting in a complicated pattern in the space-time domain. The manifestation of the physics is clearly seen from Fig. 1 in [4] in which the pressure at the lowest point is presented. Pressure spikes up to 880 bar are computed. The frequency of 750 Hz is characteristic for the particular lower head geometry. The pressure waves are the reason for small- and large-scale melt motion. The large-scale motion causes surface instability, which is partially resolved by the code numeric. The smallscale instabilities are partially resolved by the fragmentation models. Just after the interpenetrating of some fragmented melt and water occurs, the triggering happens. The local triggers are inherently predicted by the code. No artificial trigger is applied. The system possesses its internal capabilities for triggering explosions. 14.1.2.3 Material relocation Case 5 In this case the melt-water contact starts at the corner of the relocated LH - see Fig. 14.5. Then the thoroidal pressure wave presses the melt in the corner downwards. Due to mass conservation, the melt erupts in the symmetry axis. The thoroidal pressure wave accelerates the water upward in the pit. Thus we observe that the path of the melt outside the vessel is practically blocked and melt relocation into the containment due to this event is not possible. The upward melt relocation does not have enough momentum to destroy the upper RPV head. Losing its momentum the melt drops down and starts to gather in the lower head. The large amount of
14.1 Melt-water interactions
0
0.1
0.6
0.2
0.8
723
0.4
1.0 s
Fig. 14.5 An IVA5 simulation of PWR ex-vessel melt-water interaction. Case 1: Lower head unzipping and relocation in cavity partially filled with water. No water is available in the RPV. Material relocation after melt-water interaction
the water initially above the melt level is depleted. The remaining water will cyclically receive portions of melt due to melt sloshing and be depleted after one or two additional cycles. The most jeopardized region is the pit wall around the region between the LH and the remaining RPV. The peripheral melt-water interaction in this case reflects the melt into the internal region of the reactor, thus limiting the melt depletion in the containment. Case 6 This case is characterized by more violent interaction than in case 5. The inertia of the water holds the water for some time in an intensive unstable contact with the melt. The water above the melt serves as a constraint for pressure build before the expansion phase. The contact surface is much larger than in case 5. That is why the thermal energy transferred into evaporation is much larger than in case 5. The melt sloshing due to the instability, as previously described, provides additional penetration of melt into the water due to surface instabilities considerably accelerated by the local thermal interactions. It seems that the most intensive interaction starts at one
724
14 Visual demonstration of the method
half of the LH melt surface radius. This is evident from the pressure-time histories plotted in [4] for three different points over the radius at the level of the initial melt surface. The pressure wave rises first in the corner and then propagates towards the axis of the RPV being strongly amplified at a half of the radius. We observe two interactions, the first within the first 0.015 s being much shorter than the second.
Fig. 14.6 An IVA5 simulation of PWR ex-vessel melt-water interaction. Case 2: Lower head unzipping and relocation in dry cavity. Water is available in RPV. Material relocation after melt-water interaction
14.1 Melt-water interactions
725
The interaction is so violent that within about 0.2 s all the water is 10 m above its initial state - see Fig. 14.6. The water fragments during the flight and reaches the upper head in a droplet form. The impact with the upper head leads to the pressure increase. This impact does not lead to upper head failure and missile formation. The explosion disperses not only the water but the melt too. After about 0.8 s the water falls down and interacts with the melt again. The kind of interaction is very interesting: It resembles two clouds of different materials penetrating into each other. Intensive evaporation causes local pressure to again increase, which again accelerates the remaining water upwards. In general this mode of material relocation possesses some potential for depleting some melt into the containment. Again as in case 5 the most jeopardized region is the pit wall around the region between the LH and the remaining RPV. In [4] the pressures in this region as a function of time are shown. The maximum pressure at the vertical wall inside the pit is about 17.5 bar (triangle, rise time 0.03 s, decrease time 0.15 s) decreasing as one moves upwards to 12.5 bar. Case 7 is the same as case 6 except the initial pressure which was 20 bar. In the attached move to case 7 the comparison between large scale relocation of the lower head and small relocation of the lower head is demonstrated. Small scale relocation of the lower head leads to considerable low melt dispersion outside the vessel. One of the important conclusion drawn in [4] from studying cases 5 and 6 is that no significant loads result in this case for the RPV upper head. Missile formation jeopardizing the containment integrity is not possible. 14.1.2.4 Model elements addressed in cases 5, 6 and 7 1. 2.
The same model elements as in cases 1 to 4; In addition, the driving instability is not only the gravitational acceleration as in cases 1 to 4 but the deceleration due to the impact of the lower head with the bottom.
This case is extraordinarily complicated for the numerics and is recommended as a test for numerical models that will be developed in the future. 14.1.2.5 Available files on CD figs_14.05_14.06.gif, figs_14.05_14.06.html case_14.07.gif, case_14.07.html 360305"Ecugu":"vq"32" In [7] three-dimensional cases of melt-water interactions are presented. We postulate two cases 8 and 9 in which a molten sea is formed within the destroyed core of a modern PWR. In both cases the process is modeled as three dimensional. We make use of the existing symmetry and consider only 45° of the
726
14 Visual demonstration of the method
horizontal cross section. Both cases differ from each other in the elevation of the molten sea and in the side of the release slide as follows: Case 8: 15° slide, about 100 t of melt pool, bottom position. We call this case inertialess release. Case 8: 5° slide, 153 t of melt pool, elevated position. We call this case inertial release. While in case 8 the melt cannot reach considerable penetration velocity because of the short distance to the water pool, in case 9 the melt is considerably accelerated. This causes considerable differences in the physical behavior of the system. Note that in the case 7 the release is considered to happen simultaneously from 8 symmetric positions in the RPV and in case 8 from 4. 14.1.3.1 Case 8 Figure 14.7 shows the initial conditions in a horizontal plane crossing the molten pool. We see also the radial discretization used in the computational analysis.
Fig. 14.7 IVA5 PWR in-vessel melt water interaction analysis: Case 8, side failure of the heavy reflector, molten pool at the lowest position. Initial conditions and geometry in a plane crossing the molten pool
0
0.36
0.46
0.88
1.11
1.66
1.97
2.54
Fig. 14.8 Material relocation as a function of space and time
2.99 s
14.1 Melt-water interactions
727
In Fig. 14.7 the material relocation as function of time and space is presented. The discretization in the vertical plane used for the computational analysis is also visible. Reading Fig. 14.8 together with the pressure histories presented in [7] we realize the following behavior: Again as in the lower support failure cases 1 to 4 the processes are cyclic starting with small pressure increase leading to small expansion, collapse and steam explosion due to entrapment, characterized by the second sharp pressure peak. The explosion causes material relocation. The descending melt gathers again in the pool and causes a third violent pressure excursion producing the largest pressure inside the RPV. Whilst in cases 1 to 4 the water is completely depleted after 3 s in case 7 from the initial amount of about 15 t, 11 t are steel in the RPV after the third second. We observe that the inertialess side penetration of melt into the water in the lower head is very difficult and causes very low water depletion. The reverse injection of the water-steam mixture into the molten pool is possible due to the assumed large release cross section of the slice. This is the reason again for large-scale melt dispersion and depletion. After the third second from the initial mass of about 100 t, 50 t are depleted into the primary circuit. Note that there are three different geometrical regions: lower head, down comer and the empty space inside the core region. The melt is being dispersed through all of these regions. A coherent explosion influencing all regions in a sense that effective chain fragmentation happens simultaneously is impossible. The so called premixing mass that is the melt mass being in bubble three-phase flow is of particular interest. Only in bubble three-phase flow we have the most effective heat transfer mechanism from melt to liquid. During the all transient maximum 120 kg of melt are occasionally being in bubble three-phase flow. Thus, again we conclude that there is not enough available energy as a pressure source which can cause failure of the upper head endangering the containment integrity. 14.1.3.2 Case 9 Figure 14.9 shows the initial conditions in a plane crossing the molten pool.
Fig. 14.9 IVA5 PWR in-vessel melt-water interaction analysis: Case 9, side failure of the heavy reflector, molten pool at the elevated position. Initial conditions and geometry in a plane crossing the molten pool
728
14 Visual demonstration of the method
Fig. 14.10 Material relocation as a function of space and time
14.1 Melt-water interactions
729
We see also the radial discretization and the three angular sectors used in the computational analysis. Figure 14.10 shows the discretization in the vertical plane used for the computational analysis. In Fig. 14.10 the material relocation as function of time and space is shown. The material relocation is presented for three different planes. The first one is the vertical plane of the failure slice. The second one is the outermost vertical plane. The third plane is in fact parallel to the cylindrical wall of the down-comer - the outermost layer. We observe here very complex three-dimensional flow. The melt jet is disintegrated shortly after impacting the RPV wall. After the first melt reaches the lower head, the sloshing of the water in the failure plane is so strong that the dispersed melt penetrates the water very easily in the non-failure planes - see Fig. 14.10. As seen from Fig. 14.10 a large space of the down-comer is occupied by dispersed melt-water-steam flow. If one observes the movies produced with this material, one recognizes a concentration water droplet waves cyclically passing the dispersed melt. Because of the acceleration of the falling melt the water expulsion and later interactions do not force elevation of the melt. Thus, the depletion of the melt into the primary circuit is limited in this case. The pressures curves in [7] show very noisy interaction in the lower head. The lower head is filled with melt. Again as in case 8, there are three different geometrical regions where melt and water coexist: lower head, down comer and the empty space inside the core region. This melt mass is dispersed through all of these regions. A coherent explosion influencing all regions in a sense that effective chain fragmentation happens simultaneously is impossible. There is a maximum of about 1.5 t premixed melt mass. We see that from all analyzed situations 1 to 7 this is the one with the largest damage potential. The large amount of potentially explosive mixture is in fact found in the lower head. The mixture in the down-comer is rather dispersed. 14.1.3.3 Case 10 In this case we consider - 45° slice, about 120 t of melt pool, bottom position. In cases 8 and 9 we have considered melt release from the side of the heavy reflector having very pessimistic initial conditions of simultaneous release from 4 or 8 positions. In this section we consider release from one spot located at the bottom of the molten pool as presented in Fig. 14.11. As can be seen from Fig. 14.11 and Fig. 14.12, the process is asymmetric and has to be considered as a complete threedimensional problem. Geometry: The calculation has been performed in a three-dimensional cylindric geometry. The IVA computational model consists of 16 x 8 x 69 = 8 832 computational cells. These computational cells are denoted with indices (i, j, k), where i stands for the cell index in radial, j for azimuthal and k for axial direction. Schematic plots of the discretization are given in Figs. 14.11 and 14.12. Note that the lowest cell indices in all three directions have the number 2.
730
14 Visual demonstration of the method
Fig. 14.11 a, b, c IVA5 model of case 9. Discretization scheme for case 10 in two vertical planes: a) j = 2, b) j = 6
Fig. 14.12 Cross sections at different heights: a) k = 26, b) k = 62, c) k = 70
14.1 Melt-water interactions
731
Initial conditions: The initial conditions are specified as follows: The initial system pressure is assumed to be 3 bar, the system temperature is equal to the corresponding saturation temperature. We assume about 120 t of oxidic melt inside the heavy reflector and about 17 t of saturated water inside the lower head (see Fig. 14.13). The remaining part of the computational domain is filled with air.
Fig. 14.13 Initial conditions for location of melt and water. a) Cross section in r-z plane. b) Surface close to the internal wall of the vessel
Results: The computational results are presented as time functions for pressures at different positions in [7], as maps of the material relocation, Figs. 14.14 to 14.17, and as time functions of the total mass being in bubbly flow in [7]. We observe from the pressure time histories that the first contact of the melt with water happens after about 0.15 s. It produces a pressure peak of about 5.2 bar. Thereafter melt intermixing with water takes place for up to about 1 s. At that time explosive events lead to strong local pressure changes with time, followed cyclically by two other weaker interactions. Pressure peaks of up to 12.4 bar are observed. The comparison of the pressure time histories with the material relocation pictures reveals that, after the first melt-water contact happens, the interactions locally produce such a pressure that causes melt accumulation immediately after the release. Just after the pressure release the accumulated melt drops down into the water and creates an impact intermixing which leads to explosive interactions. The result is acceleration of the water into the depletion cross sections. The local pressure increase causes also a stop of the melt release. Moreover, it blows multi-phase mixture into the melt through the opening cross section in the heavy reflector. The melt contains a certain amount of water. This is the reason for internal interaction inside the molten pool leading to local dispersion. Due to the limited amount of water content in the mixture the interaction is not strong enough to accelerate the melt up to the top. After some time the melt settles down and starts again to flow continuously through the failure cross section (see Fig. 14.14).
732
14 Visual demonstration of the method
Fig. 14.14 Material relocation at different times (planes j = 2 and j = 6)
14.1 Melt-water interactions
0
1.0
0.15
1.2
0.3
1.5
0.5
0.9
0.95
2.0
2.5
3.0 s
733
Fig. 14.15 Material relocation in the lower head at different times (plane j = 2)
The asymmetric behavior of melt and water can be clearly seen in Fig. 14.16, where two planes are shown, one through the sector containing the leak and the other through the diametrically opposed sector.
0
0.15
0.3
0.5
734
14 Visual demonstration of the method
0.9
1.0
1.1
1.2
1.5
2.0
2.5
3.0 s
Fig. 14.16 Material relocation in the peripheral direction at different times (plane i = 15)
14.1 Melt-water interactions
735
Fig. 14.17 Material relocation – melt and water iso-surfaces, 1.07, 1.5 s
How the water behaves, is illustrated by comparing Figs. 14.15, 14.16 and 14.17. On the side of the melt ingression the melt-water interaction pushes the water away. The intermixed masses are only about 300 kg, so that the cyclically released mechanical energies are not sufficient for complete water depletion from the reactor pressure vessel. As seen from Fig. 14.16 a massive bulk of water is subject to azimuthal concentration waves. The water entrainment happens mainly due to fragmentation and mechanical steam-water interaction. The asymmetric geometry, as is clearly seen from Fig. 14.17, causes asymmetric behavior in the down comer. No more than 1.3 t of melt is able to release its energy. Finally, the pressurization events are not amplified but are essentially damped while propagating up to the upper head - as can be seen in the pressure time histories presented in [7]. 14.1.3.4 Model elements addressed in cases 8 to 10 1. 2.
The same model elements as in cases 1 to 7; In addition, the numerical method is challenged for all three dimensions.
Again these cases are extraordinarily complicated for the numeric and are recommended as a test for numerical models that will be developed in the future.
736
14 Visual demonstration of the method
14.1.3.5 Available files on CD fig_14.08.gif, fig_14.08.html fig_14.09.gif, fig_14.09.html fig_14.10.gif, fig_14.10.html fig_14.10.gif, fig_14.10.html fig_14.10.gif, fig_14.10.html 360306"Ecugu"33"cpf"34" In [7] two cases are presented for ex-vessel melt spreading into a compartment with a flat horizontal surface, one without a water layer and one with a water layer. 14.1.4.1 Case 11 We simulated gravitational sloshing of 300 t of non-modified melt at 3000 K in a dry compartment comprising a 110o sector with horizontal spreading surface. Note that for this melt the dynamic viscosity of 0.0044 kg/(ms) has no influence on the spreading process, as in this particular case the spreading is inertia dominated. The melt enters the computational region at the axis of the spreading compartment. As the problem is symmetric, only half of the geometry is simulated. Fig. 14.18 shows the dicretization net and the surface dividing volumetric concentration of the melt being less and larger than 0.005. The result shows that the spreading is turbulent and heterogeneous. Two symmetric vortices are formed within 20 s. In the center of the vortices there are regions which are not reached by the melt in this initial period. The melt covers the total area in about 30 s. Gravitational surface waves travel for some time after this.
Fig. 14.18 An IVA5 prediction of the dry melt spreading dynamics in a EPR spreading compartment. Iso-surface of the melt volume fraction with overlaid melt temperature at the surface
14.1 Melt-water interactions
737
Fig. 14.18 (Cont.)
14.1.4.2"Case 12 In this case we analyze the spreading using similar initial and boundary conditions as in the previous case with the only difference that the floor is covered with 1 cm water and the geometry is slightly different.
Fig. 14.19 IVA5 discretization model for simulation of melt spreading in presence of water
Fig. 14.20 An IVA5 prediction of the wet melt spreading dynamics in a EPR spreading compartment. Iso-surface of the melt volume fraction with overlaid melt temperature at the surface. Iso-surface of the water volume fraction with overlaid water temperature at the surface - blue. Spreading of melt in presence of thin water layer (1cm). Parameter: Time 0, 2, 3.8, 6.2, 8, 10, 14.4, 16.8, 17.8, 18 s
738
14 Visual demonstration of the method
Fig. 14.20 (Cont.)
Figures 14.19 and 14.20 present the melt-water interaction in this case. The melt and the water are presented with iso-surfaces dividing volume fractions less than and larger than 0.5%. In [7] the pressures at two different places in the spreading compartment are presented. The pressure spikes are characteristic for inertial water entrapment. The discretization used in this case is not fine enough to resolve such events with large confidence. 14.1.4.3"Model elements addressed in cases 11 and 12 1. 2. 3.
The same model elements as in cases 1 to 10; In addition, the numerical method is challenged for modeling of stratification of multiple materials. The iso-surfaces possess as an attribute the temperature of the constituent. The thermal radiation and temperature reduction of the surface are clearly visible.
14.1 Melt-water interactions
739
14.1.4.4"Available files on CD fig_14.18.gif, fig_14.18.html fig_14.20.gif, fig_14.20.html 360307"Ecug"35" Figures 14.21 and 14.22 demonstrates a vessel filled with inert gas, hydrogen and molten material that is discharged in the container air atmosphere through a defined opening at the lower heat. The pressure in the vessel and in the stagnation point of the jet is presented also in the Fig. 14.22. 14.1.5.1 Model elements addressed in case 13 1. 2.
Fragmentation and the associated thermal interaction between the inert gas and the molten material; Burning of hydrogen during the process.
Fig. 14.21 A high pressure discharge of melt into a cavity, geometry (Sandia National Laboratory), initial conditions, Kolev (2000b)
740
14 Visual demonstration of the method
Fig. 14.22 A high pressure discharge of melt into a cavity
14.1.5.2 Available files on CD fig_14.22.gif, fig_14.22.html 360308"Ecug"36" In [9] severe accident control for a large-scale boiling water reactor through invessel melt retention by external reactor pressure vessel cooling is considered. Side release of melt from a postulated molten pool in the core region is analyzed.
14.1 Melt-water interactions
741
Fig. 14.23 Material relocation in the reactor pressure vessel as a function of time and space. 0.18 m² melt release cross section. Legend: blue – water, red – melt, transparent – gas
We assume a vertical opening of the core shroud wall with the height of 0.9 m and horizontal size of 20 cm which gives 0.18 m² cross section. The discretization contains (8x6x29)= 1392 cells for the half of the reactor vessel. We assume 1.6 t
742
14 Visual demonstration of the method
steam within the rector vessel with temperature 419.15 K, 124 t water with temperature 417.75 K, 186 t melt with temperature 3000 K – see Fig. 14.23. The initial pressure is 4.1 bar. We assume pressure 4.1 bar at the 6 safety valve lines that are open and remain open. 14.1.6.1 Results of the computational analysis We present in Fig. 14.23 the material relocation as a function of the time and space. In [9] the pressures as functions of time in the down-comer periphery below the initial water level, the pressures as functions of time in the lower head at different positions, and the pressures as functions of time in the upper head at different positions are presented. Comparing the material relocation with the pressure history we reveal several interesting phenomena. The process requires 200 s to come to the natural end at which there is no more water to interact with the melt any more. This is much longer than the characteristic time of 1 to 3 s for meltwater interactions in PWR’s – see in all previously discussed cases During this time the melt releases a considerable part of its internal energy as seen from the color of the melt representing the melt temperature. The melt reaches actually the lower head in a form of solid particles. For about 70 s the level of the pressure is around 12 bar. During this time there is pulsating water exchange between the lower head and the down-comer. Melt-water interactions in the down comer cause considerable water dispersion waves. Water cyclically comes in contact with the melt and is then being repulsed. Short living pressure pulses up to 25 bar are also shown in [9]. Limitation of the present day computer power is the main reason not to attack this complicated multi-phase problem with the same resolution as is now standard for single-phase flow. Grid dependence is investigated increasing the number of the cells up to (12x12x29) = 4176 for half of the reactor vessel. We found that the pressures satisfy the findings that the uncertainty in the pressure prediction is within 50% of the pressure increase. The main conclusion of studding this particular case made in [9] was that the melt-water interaction causes pressure spikes below 30 bar maximum pressure which again having the uncertainties in mind with which such computations are associated [9] is of no concern for the integrity of the pressure vessel. Therefore there is no danger that the external cooling strategy will not start properly. 14.1.6.2 Model elements addressed in case 14 1. 2.
All elements as in the all previous cases; The main problem associated with this analysis is the long duration of the melt-water interaction compared to all previous cases and the need to resolve pressure spikes as in the cases of short time interactions.
14.2 Pipe networks
743
14.1.6.3 Available files on CD fig_14.23.gif, fig_14.23.html
3604"Rkrg"pgvyqtmu" 360403"Ecug"37" This case is in contrast with all previous cases. It is a simulation of processes inside a complicated network consisting of pipes, valves etc. – see Fig. 14.24 [8]. The heat exchanger has as primary medium water at 320°C at about 160 bar pressure and as secondary medium water with 50°C at about 6 bar. A break inside the heat exchanger is simulated. Flashing water enters in the secondary side and creates non-stable pressure increase. The first 0.2 s are characterized by strong condensation oscillation in the secondary side. Complicated interactions between the valves and the break are extremely challenging to the solution method.
Fig. 14.24 Pipe break in a high-pressure heat exchanger. Pressure as a function of time
744
14 Visual demonstration of the method
Fig. 14.24 (Cont.)
14.2.1.1 Model elements addressed in case 15 1. 2. 3. 4.
Two-phase flow in complex pipe networks; Networks containing valves interacting with the flow; Completely opposite physical phenomena like flushing and condensation shocks happen in the same network; Constitutive models for flushing and condensation shocks.
14.2.1.2 Available files on CD fig_14.24.gif, fig_14.24.html
14.3 3D steam-water interaction
745
3605"5F"uvgco/ycvgt"kpvgtcevkqp" 360503"Ecug"38"
Fig. 14.25 a) 3D test section simulating full scale PWR for hot leg water injection in steaming core. b) IVA6 geometry model for UPTF-Test 26 Run 230 simulation. c) Water volume concentration and steam velocity as a function of space at different times. Velocities at the entrance of the main circulation pipe for the corresponding times
Figure 14.25 a) and b) presents a technical facility designed for studying of mechanical steam-water interaction in complicated geometry – see [10]. A water jet
746
14 Visual demonstration of the method
is injected from the side of the upper part, and steam is injected inside the vessel on a specified horizontal level through a part of the cross section. Figure 14.25 c) presents an IVA6 simulation of this process. 14.3.1.1 Model elements addressed in case 16 1. 2.
Geometry representation using the concept of the heterogeneous permeabilities; Fragmentation and coalescence dynamics, represented here with conservation equations for a given class of particle number densities, etc.
14.3.1.2 Available files on CD fig_14.25.gif, fig_14.25.html
3606"Vjtgg"fkogpukqpcn"uvgco/ycvgt"kpvgtcevkqp"kp" rtgugpeg"qh"pqp/eqpfgpucdng"icugu" 360603"Ecug"39" Controlling the pressure of pressurized system is one of the most important safety tasks of the design engineer. Usually safety and relieve valves helps to release working material if for some reason the pressure increases. In many cases the working material is not released to the atmosphere but to a especially designed for this purpose vessel. For systems working with steam this may happen in water vessels in order to reduce dramatically the volume of the steam. The vessels are usually filled partially with water and partially with inert gas. Knowing the phenomena in such a release process allows designing properly the system. Let consider one such case, see Fig. 14.26. The biggest challenge here is the very sensitive condensation process at bubble and droplet surfaces that changes dramatically. We open the valve over about 22 seconds. Thereafter the valve is closed. Several interesting processes can be recognized from the attached move. Initially the pressure in the pressure dome increases up to 10 bar in order to overcome the inertia of the water and to remove it from the dome. In front of the water we observe haw the initial volume of the inert nitrogen was dramatically reduced. The water first penetrates in the release vessel followed by the inert gases. This causes increasing the water level. Due to the intensive input of the mechanical energy the system is so much turbulized that within a few seconds all the vessel is occupied by rotating two phases, two component mixture. The condensation of the steam happens in the immediate neighborhood of the distribution nozzles. Large bubble oscillation in the vessel causes increasing and decreasing resistance to the steam resulting in some macroscopic pressure oscillations. After the valve is stopped, the pressure in the dome decreases exponentially. After reaching a values lower then the vessel pressure
14.4 Three dimensional steam-water interaction in presence of non-condensable gases
747
some waters penetrates back into the distribution system. Because there is only pure steam in the dome the condensation creates a vacuum of 0.2 bar. The inverse pressure difference causes change of the flow direction in the nozzles. The stratification process at that moment in the vessel has just started and there is enough rotation energy keeping the mixture none stratified. The result is the two-phase penetration back into the dome. But the gas inside the vessel was nitrogen. Therefore nitrogen is transferred in this way back into the dome reducing the non wished vacuum production in the dome.
Fig. 14.26 Release of saturated steam (p = 175 bar) into vessel partially filled with 50°C water through 40 cm2 cross section valve
14.4.1.1 Model elements addressed in case 17 1. 2. 3.
Geometry representation using the concept of the heterogeneous permeabilities, permeabilities being a time functions; Fragmentation and coalescence dynamics, represented here with conservation equations for a given class of particle number densities, etc.; Steam condensation (bubbles, droplets) in subcooled liquid;
748
4. 5.
14 Visual demonstration of the method
Steam condensation (bubbles, droplets) in subcooled liquid in presence of non-condensable gases; Dynamic flow regime transitions.
14.4.1.2 Available files on CD fig_14.26.gif, fig_14.26.html
3607"Vjtgg"fkogpukqpcn"uvgco"rtqfwevkqp"kp"dqknkpi"ycvgt" tgcevqt" 360703"Ecug"3:" Modern design of boiling water reactors requires analysis of the steam quality produced by the nuclear reactor. We give here an example from a design case study in which 1/4th of the modern reactor vessel was presented by about 10000 cells. The thermal heat release in the reactor core is 3D-non uniform. The computation is performing starting with and arbitrary but meaningful initial state until reaching the steady state. The results are presented in Fig.14.27 for the horizontal plane being at the outlet of the core. Such studies are valuable tools for optimizing design geometry. 14.5.1.1 Model elements addressed in case 18 1. Geometry representation using the concept of the heterogeneous permeabilities; 2. Fragmentation and coalescence dynamics, represented here with conservation equations for a given class of particle number densities, etc.; 3. Dynamic flow regime transitions. 4. Thermal structure-flow interactions. 5. All forced convection heat transfer regimes: forced convection, subcooled boiling, saturated forced convection boiling, departure of nucleate boiling etc. 6. All kind of interaction of three velocity field with each other. 7. Appearance and disappearance of velocity fields. 14.5.1.2 Available files on CD fig_14.27.gif, fig_14.27.html
References
749
Fig. 14.27 Volume fraction of steam in a boiling water reactor - top view into 1/4th of the reactor vessel
Tghgtgpegu" 1.
Kolev NI (October 15-16, 1996) Three fluid modeling with dynamic fragmentation and coalescence - fiction or daily practice? 7th FARO Experts Group Meeting Ispra; (5th-8th November 1996) Proceedings of OECD/CSNI Workshop on Transient thermal-hydraulic and neutronic codes requirements, Annapolis, MD, USA; (June 2-6, 1997) 4th World Conference on Experimental Heat Transfer, Fluid Mechanics and Thermodynamics, ExHFT 4, Brussels; (June 22-26, 1997) ASME Fluids Engineering Conference & Exhibition, The Hyatt Regency Vancouver, Vancouver, British Columbia, CANADA, Invited Paper; (May 22-24, 1977) Proceedings of 1997 International Seminar
750
14 Visual demonstration of the method
on Vapor Explosions and Explosive Eruptions (AMIGO-IMI), Aoba Kinen Kaikan of Tohoku University, Sendai-City, Japan 2. Kolev NI, Chen T, Kollmann T, Schlicht G (June 8 - 12, 1998) Visual multi-phase flow analysis, Third International Conference On Multiphase Flow, Lion, France 3. Kolev NI (17 December 1998) In-vessel melt-water interaction caused by core support plate failure under molten pool, Part 2: Analysis, 5th MFCI Project Meeting, Forschungszentrum Karlsruhe, Germany; (October 3-8, 1999) Proc. of the Ninth International Topical Meeting on Nuclear Reactor Thermal Hydraulics (NURETH-9), San Francisco, California, Log. Nr. 316_2; Kerntechnick, vol 64 no 5 pp 278-283 4. Kolev NI (17 December 1998) Ex-vessel melt-water interaction caused by symmetric lower head unzipping within the reactor pit, 5th MFCI Project Meeting, Forschungszentrum Karlsruhe, Germany; Slightly modified in (April 19-23, 1999) Proc. of the 7th International Conference on Nuclear Engineering, Tokyo, Japan, ICONE-7361; Extended version in (October 3-8, 1999) Proc. of the Ninth International Topical Meeting on Nuclear Reactor Thermal Hydraulics (NURETH-9), San Francisco, California 5. Kolev NI (5-10 September 1999) Applied multi-phase flow analysis and its relation to constitutive physics, Proc. of the 8th International Symposium on Computational Fluid Dynamics, ISCFD ‘99 Bremen, Germany, Invited lecture; (April 2000) Japan Journal for Computational Fluid Dynamics, vol 9 no 1 pp 549-561; (May 20-25, 2001) 13th School-Seminar of Young Scientists and Specialists, Physical Principles of Experimental and Mathematical Simulation of Heat and Mass Transfer and Gas Dynamics in Power Plants, Saint-Petersburg, Russia 6. Kolev NI (October 3-8, 1999) Verification of IVA5 computer code for melt-water interaction analysis, Part 1: Single phase flow, Part 2: Two-phase flow, three-phase flow with cold and hot solid spheres, Part 3: Three-phase flow with dynamic fragmentation and coalescence, Part 4: Three-phase flow with dynamic fragmentation and coalescence – alumina experiments, Proc. of the Ninth International Topical Meeting on Nuclear Reactor Thermal Hydraulics (NURETH-9), San Francisco, California 7. Kolev NI (April 2-6, 2000a) Computational analysis of transient 3D-melt-water interactions, Proc. of the 8th International Conference on Nuclear Engineering, Baltimore, Maryland USA, ICONE-8809; Also in (May 22-25, 2000) CFD 2000 in Trondheim Norway; Also in abbreviated form in (November 5-11, 2000) Symposium on “Dispersed Flows in Combustion, Incineration and Propulsion Systems”, ASME International Mechanical Engineering Congress & Exposition, Orlando, FL 8. Kolev NI (June 11-15, 2000b) Needs of industrial fluid dynamics applications, Invited lecture, 2000 ASME Fluids Engineering Division Summer Meeting (FEDSM), Industry Exchange Program, Sheraton Boston Hotel, Boston, Massachusetts 9. Kolev NI (April 2-12, 2001) SWR 1000 Severe accident control through in-vessel melt retention by external RPV cooling, 9th International Conference on Nuclear Engineering, Nice, France 10. Kolev NI, Seitz H and Roloff I (May 27–June 1, 2001) Hot leg injection: IVA 3D Versus 1D Three Velocity Fields Modeling and Comparison with UPTF 26 Run 230 Experiment, ICMF-2001, CD-Proceedings of the 4th International Conference on Multiphase Flow, New Orleans, Louisiana, USA; In abbreviated form (September 24-28, 2001) Proc. ExHFT-5, 5th World Conference on Experimental Heat Transfer, Fluid Mechanics and Thermodynamics, Thessalonica, Greece
Index
absolute coordinates 556 absolute temperature 176, 231, 413 acoustic waves 410 added mass force 82, 504 adiabatic temperature of the burned gases 298 Air containing microscopic impurities like dust 118 Al2O2 – water 431 anergy 315 angle between two nonzero vectors 645 angles between the unit covariant vectors 670 anisotropic forces 84 aperiodic 362, 372 aperiodic case 364 approximating the contravariant base vectors 680 arc length 43, 669 arc length parameter base point 641 arc length vector 43 area of parallelogram 655 area of union of two triangle plane regions 655 Arrhenius-law 294 associative and distributive lows for the cross product 654 asymptotic 332, 363 asymptotic case 374 August equation 147 averaging 10 axial pump 389 Back pressing valves 357 baro-diffusion 20 bcoefficients 606 Bernulli equation 571 block Newton-Raphson method 511 Block-Gauss-Seidel method 581 Board 408
boiling at the wall 273 Boussinesq hypothesis 59, 61 bulk pressure 51 canonical form 179, 203, 475, 477, 479, 527 Carnot coefficient 313 Carnot cycle 147, 319 Cartesian vector in co- and contravariant coordinate systems 668 Cauchy’s lemma 42 center of mass velocity 224, 339 centrifugal force 570 centrifugal pump 378, 392 centrifugal pump drive model 380 centrifugal pumps 378, 392 centroid of triangle 637 centroid of union of non overlapping plane regions 637 centrufugal forces 104 Chapman-Jouquet condition 413, 417, 426, 428, 430 characteristic equation 476 characteristic points 556 characteristic velocity 478 chemical equilibrium 140, 146 Chemical equilibrium conditions 300 Chemical potential 136 chemical reaction in gas mixture being in closed volume 292 Clapeyron 147 Clapeyron‘s equations 147 Clausius 147, 148, 167, 176, 279 closest possible packing of particles 26 coalescence 27 coefficient of performance 319 cofactor of the Jacobian matrix 706 collision 34 Combustion in pipes with closed ends 418
752
Index
combustion kinetics 294 compression work 316 concentration equations 514, 518, 594 condensation 268 condensation sources 267 conservation equations in general curvilinear coordinate systems 445 conservation of energy 174 conservative form, 449 conservative form of the energy conservation equation 222 conservative form of the mass conservation equation for multi-phase flows 448 conservative scheme 525 constrained interpolation profile (CIP) method 525 continuum-sound waves 409 continuum surface force 53 contraction number 330 contravariant base vectors 667 contravariant metric tensor 678 contravariant vectors 446, 597, 679 control volume 4, 174 convection + diffusion velocity 492 coolant entrainment ratio 439 Coriolisis force 104 Courant and Hilbert 477 Courant, Friedrichs and Levi 519 covariant base vectors 664 covariant metric tensor 670 critical flow 335 Critical flow 480 critical mass flow rate 342, 345 curl 676, 683 curvature 35 curvature of a surface 648 curvature of a surface defined by the gradient of its unit normal vector 649 curvature of space curve 643 curvilinear coordinate system 445, 587 Cylindrical coordinates 696 D’Alembert 174 Dalton 117 Dalton’s law 2, 19, 119, 196, 223, 228, 299 debris quenching 434 deformation 329 deformation of the mean values of the velocities 236, 254, 255
dependent variables vector 181 deposition 28 deposition rate 93 derivative of a position vector defining a curve 639 derivative of a position vector defining a line 639 derivatives 712 derivatives along the normal to the curvilinear coordinate surface 687 derivatives along the tangent of the curvilinear coordinate lines 687 derivatives for the equations of state 115 derivatives of a function along the unit tangent vector 653 determination formula for 654 Detonability concentration limits 297 detonation adiabatic 414 detonation in closed pipes 418 detonation in perfect gases 414 detonation wave 431 Detonation waves in water mixed with different molten materials 431 diagonal dominance 508 diagonal pump 389 diagonalizable matrix 659 diagonally dominant 610 differential form 652 differential form of equation of state 124 differentiation rules for time derivatives 642 diffusion Laplacian 449, 699, 703 diffusion Laplacian of tensor 686 diffusion velocity 19, 224 directional derivatives 652 Dirichlet boundary value problem 704 discharge of gas from a volume 285 discontinuum-shock waves 430 discretization 488 discretized concentration conservation equation 593 discretized concentration equation 611 discretized mass conservation 611 discretized momentum equation 609 dispersed flow 56, 79 dispersed interface 77 dissociated steam 417 divergence 674, 675, 683 divergence of tensor 686
Index divergence theorem 674 divergence theorem for curvilinear coordinate system 675 donor cell concept 487 donor cell principle 500 donor-cell 594 drag force 617 dual vectors 679 dumping time constant 362 dyadic product of two vectors 658 eddy conductivity 237 effective gas constant 283 eigenvalues 69, 203, 361, 476, 659 Eigenvalues and critical flow 480 eigenvectors 477, 659, 660 elasticity modules 330 Elbows 558 elliptic 477 elliptic grid generation systems 703 energy concept 131, 135, 151, 154 energy conservation 173 energy conservation equation 175 energy equation 176 energy jump condition at the interfaces 267 energy principle 216 enthalphy 216 enthalphy concept 150 enthalphy diffusion coefficient 242 enthalpy concept 132, 154 enthalpy equation 225 entrainment 28 entropy 131 entropy change due to the mixing process 136 entropy concept 116, 135, 151, 154, 202, 264 entropy diffusion coefficient 240 entropy equation 178, 194, 198, 202, 213, 230, 232 equation of state 177 equations for plane in space 644 equations for tangent planes and normal lines 646 equilibrium between the liquid and its vapor 146 Equilibrium dissociation 298 equilibrium mixture of liquid and solid phase 164 Euler 1, 174
753
Euler equations 344 Eulerian description 174 evaporation 268 evaporation sources 267 exact differential form 652 exergetic coefficient of performance 319 exergy 311 explosivity of melt-water mixtures 407 Fe – water 408, 430, 431, 441 Fick 19 Fick‘s law 19, 223 field entropy conservation equations 453 field mass conservation equations for each component 449 finite difference method 598 finite volume method 597 first fundamental metric identity 448, 676, 683, 695 first order donor cell 490 first order donor-cell finite difference approximations 490 first-order donor-cell 594 five-field concept 425 flashing 272 flow patterns 2 flux concept 466 form drag 83 forth-order-polynomial spline 525 Fourier equation 407 fragmentation 25 free enthalpy 139 free turbulence 237 frequently used curvilinear coordinate transformations 696 frictional pressure drop 340 fuel equivalence ratio 295 fugacity 144 fundamental exergy equation 316 gas phase 2 gas release due to pressure drop 118 Gases dissolved in liquid 117 Gases inside a gas mixture 117 Gauss-Ostrogradskii theorem 50 Gauss-Ostrogradskii 674 Gauss-Ostrogradskii theorem 11 Gauss-Seidel 489, 517 Gauss-Seidel iteration 489
754
Index
Gauss-Seidel method 581, 595 Gay-Lussac 177 general transport equation 12 Gibbs energy of the chemical reaction 144 Gibbs equation 197 Gibbs equation for mixtures 135 Gibbs function 139, 146 Gibbs mixture equation 197 Gibbs potential 136 Gibbs relation 176 Gibs function 139 Gouy exergy definition 316 gradient 676, 679, 683 gradient along a direction normal and tangential to a surface 648 gradient of plane defined by three points 644 gradient vector (gradient): 646 grid 589 Hagen-Poiseuille law 253 harmonic averaging 493, 606 harmonic oscillations 361, 365, 475 harmonic oscillations case 363 head and torque curves 388 heat conduction equation 477 heat input in gas being in closed volume 287 heat pump 318 Helmholtz and Stokes hypothesis 454 Henry and Fauske 344, 348, 433 high order discretization 484 high order upwinding 481 homogeneous flow 346 homogeneous multiphase flow 422, 423 homologous curve 384 Huang Hu Lin 384 Hugoniot 412 hydraulic diameter 14 hydrogen combustion in inert atmosphere 294 hyperbolic 477 ignition concentration limits 297 Ignition temperature 297 impermeable fixed surfaces 15 incompressible flow 331 increment of the arc length 669 increment of the position vector 669 inert components 2
inert mass conservation equation derivatives 580 infinite volume in curvilinear coordinate systems 672 injection of inert gas in closed volume initially filled with inert gas 287 integration procedure 517 interface conditions 273 interfacial area density 14, 53 interfacial energy jump condition 220 interfacial heat transfer 266 interfacial mass transfer 345 interfacial momentum transfer due to mass transfer 618 internal energy equation 229, 239 intrinsic field average 8 intrinsic surface average 8 intrinsic surface-averaged field pressure 51 inverse metrics 598 inverted metrics tensors 446 irregular grid’s 589 irregular mesh 590 irreversible bulk viscous dissipation 227 irreversible dissipated power 236, 254, 255 irreversible power dissipation 454 Isaac Newton 41 isentropic 286 isentropic exponent 152 isothermal coefficient of compressibility 157 isotropic convection-diffusion problem 450 iteration method 508 IVA2-method 610 Jacobi method 498, 581 Jacobian 449, 512, 666 Jacobian determinant 592, 598 Kelvin-Helmholtz stability 71, 76 kg-mole concentrations 121 Kirchhoff 148 knots 554 KROTOS 435 Lagrangian 174 Landau and Lifshitz 286, 298 Laplace 409 Laplace equations 710
Index Laplacian 676, 683, 703 Lava consisting of several molten components containing dissolved gasses 118 laws of the dot product 645 lean mixtures 431 Leibnitz rule 50, 217 lift force 91, 654 line spacing control functions 706 linear stability analysis 519 linear system of equations 580 linearized form of the equation of state 609 Linearized form of the source term for the temperature equation 265 linearized state equation 507 lines and line segments in space 638 liquid water 2 Liquid water containing dissolved gases 118 Liquids dissolved in liquid 117 local critical mass flow rate 341 local instantaneous entropy equations 232 local volume and time average mass conservation equation 22 local volume average 8 local volume averaged 175 local volume-averaged momentum equations 42 locally monotonic behavior 524 Lomonosov 1, 174 lumped parameter volumes 273 Mach number 124, 335 macroscopic density 8 magnitude 702 main pump equation 382, 398 Marangoni effect 46, 77 mass concentrations 121 mass conservation 2, 13, 305, 500 mass conservation derivatives 578 mass conservation equation 17, 449, 546, 594 mass conservation equation for each specie inside the velocity field 326 mass conservation equation for the microscopic component 18 mass conservation equations 507 mass flow concentrations 339 mass flow vector 592
755
mass jump condition 22 mass source 14 Melt-coolant interaction detonations 407 melt-water detonation 407 mesh 590 method of characteristics 187 metrics 446 metrics and inverse metrics 665 midpoints 637 miscible components 117 mixture consisting of steam and air 165 mixture density 119, 125, 262, 505 mixture enthalpy 197 mixture isentropic exponent 263 mixture mass flow rate 339 Mixture of liquid and microscopic solid particles of different chemical substances 153 mixture of saturated steam and saturated liquid 155 Mixture of solid particles and liquid 117 mixture specific heat 283 mixture volume conservation equation 260, 505 Mixtures of non-miscible liquids 117 molar concentration 121 molecular diffusion 20 momentum conservation 56 momentum conservation equations 459 momentum equation 55, 338, 500, 550, 575 momentum equations 512, 518 momentum jump condition 48, 51 mono-dispersity 25, 30 moving coordinate system 690 multi-blocks 590 Multi-component mixtures of miscible and non-miscible components 117 multi-phase flow patterns 3 necessary condition for convergence 497 Newton 174 Newton method 139 Newtonian continuum 57 Newton-Raphson 121 Newton-type iteration method 508 nodes 589 non-condensibles 273 non-conservative form 340, 449
756
Index
non-conservative form of the concentration equations 594 non-conservative form of the energy equation 222 non-conservative form of the enthalpy equation 226 non-conservative form of the entropy equation 256 non-conservative form of the momentum equation 225 non-miscible components 118, 120, 123 non-monotonic behavior 525 non-physical diffusion 15 non-slip boundary condition 504 non-structured 2 nozzle flow 338 nozzle flow with instantaneous heat exchange without mass exchange 338 nozzles frozen flow 336 nucleation rate 27 Nukiama-Tanasawa distribution 30 number density of particles 24 numerical diffusion 15 numerical methods 481 numerical orthogonality 710 numerical solution 189 numerical solution method 458 off-diagonal diffusion terms 568 off-diagonal viscous forces 618 one dimensional three-fluid flow 325 order of the chemical reaction 294 orthogonal coordinate systems 600 orthogonal systems 710 orthogonality 671 orthotropic 695 oscillation period 362 outer iterations 518 parabolic 523 parallel vectors 654 partial decoupling 497 partial derivatives 124, 154, 674 partial differential equations 482 partial pressures 117 Partial pressures of the perfect fluid compounds being in chemical equilibrium 144 particle number density conservation equation 451
particle number density derivatives 578 particle number density equation 326, 605 particles number density equation 469 pdv-exergy 176 perfect fluid 119 perfect gas 134, 147 perfect gases 129 perpendicular (orthogonal) vectors 646 Pfaff‘s form 176 phase 2 phase discontinuity treated with CIP 529 Phase equilibrium 146 physical space 695 pipes 329, 554 pipe deformation 329 pipe library 559 pipe networks 325, 378, 554 pipe section 556 plane determined by three points 644 Poison-like equations 703 Poisson-type equation for multi-phase flows 514 polytrophic exponent 284, 288 Position vector defining a curve 639 position vector defining a line 639 potential equation 477 potential function 651 Prandtl-Kolmogorov law 259 pressure equations 505, 507 pressure equilibrium factor 143 pressure force 350 pressure-velocity coupling 587, 596 principal unit normal vector for space curve 643 principle of conservation of mass 13 propagation velocity of harmonic oscillations 475 proportions 145 pseudo gas constant 152 pseudo-canonical form 477 pseudo-exergy equation 312, 315 pump behavior 378 pump model 378 pumps 378 Rayleigh–Taylor wavelength 412 Rankine 412 Rankine-Hugoniot curve 414 Ransom 76, 385
Index reaction progress variable 142 reaction thrust 353 reaction velocity 294 reduction 557 regular grid’s 589 regular mesh 590 regular topology 590 RELAP 385 relation between the contravariant and the covariant base vectors 679 relations between the inverse metrics and the metrics 681 relative coordinates 556 relief valves 356 residual error vector 581 restatement of the conservative derivative operations 683 restatement of the non conservative derivative operations 684 Reynolds averaging 16 Reynolds stresses 56 right handed Cartesian coordinate system 637 risk potential of dispersed-dispersed systems 431 r-momentum equation 568 rotation around axes 660 saturation line 147 Sauter mean diameter 31 scalar (dot) product 643 Scarbourough criterion 498 Scot and Berthoud 408 second fundamental metric identity 449, 695 second law of the thermodynamics 259 semi-conservative form for the entropy equation 244 semi-conservative forms 243 shock adiabatic 414 shock tube 203 shock wave discontinuity 16 shock waves 409 signed integers 564 signet integers 594 single component flashing nozzle flow without external sources 343 single phase momentum equations 42 Single-component equilibrium fluid 155 single-phase test cases 107 slip 337
757
Slip model 338 slip ratio 337 slip velocity ratio 339 solid particles 2 solid phase 164 sonic velocity 263, 328, 458, 505 sound velocity 410 sound wave propagation in multi-phase flows 421 sound waves 409 space exponential scheme 522 spatial averaging theorem 10 specific enthalphy equation 241 specific enthalpy 178 specific entropy derivatives 579 specific mixture entropy of the velocity field 232 specific technical work 316 spectrum of particle size 30 speed of displacement of geometrical surface 650 spherical coordinates 700 splitting a vector 644 stability criterion for bubbly flow 98 staggered grid 484 staggered grid methods 613 staggered x-momentum equation 613 standard parametrization of the line 638 steady state flow 333 steam injection in steam-air mixture 289 Stokes theorem 44 stratified flow 65, 71, 329 stratified structure 29 strong detonation waves 416 strong pressure waves 409 structured 2 sub system network 559 successive relaxation methods 508 superheated steam 155 surface area increment 671 surface averaged velocity 16 surface force 43, 53 surface permeabilities 483 surface permeability 6 surface tension 43 Suter-diagram 386 Suter-diagrams 389, 391 system of algebraic equations 534 system of algebraic non-linear equations 509
758
Index
temperature 177 Temperature and pressure change in closed control volume 305 temperature equation 198, 265, 278, 327, 468, 494, 497, 604 temperature inversion 429, 513, 519 The heat release due to combustion 297 thermal expansion coefficient 156 thermo-diffusion 20 thermodynamic and transport properties 167 thermodynamic equilibrium 163 thermodynamic properties of water and steam 168 thermodynamic state 156 thickness of the discontinuity front 408 third low of the thermodynamics 158 thrusts acting on the pipes 350 time average 16 time averaged entropy equation 234 time dependent grid metric identity 695 time step limitation 519 topology 589 total time derivative 14 total wall force 352 transformed coordinate system 446 transformed fluid properties conservation equation 692 transformed space 664 transformed total time derivative of a scalar 695 transient forces 357 Translation 660 triple scalar or box product 656 turbulent coefficient of thermal conductivity 237 turbulent diffusion 21, 92, 236, 237 turbulent pulsations 236, 250, 251, 254, 255
vector of dependent variables 475, 481, 590 velocity discontinuity in the shock plane 412 velocity field 1 velocity field no. 2 2 velocity field no. 3 3 velocity field no.1 2 velocity of sound 159 vena contracta 360 virtual mass force 81 virtual mass term 462 viscous diffusion 600, 525, 630 viscous dissipation 227 viscous dissipation rate 252 Volcanic explosions 118 volume conservation equation 260, 458 volume elements 590 volume fraction 6 volume of hexahedron 657 volume of space enveloped by a closed surface 675 volume of tetrahedron 657 volume porosity 546, 547 volumetric mixture flow rate 264 volumetric porosity 5
unit tangent vector 642 UO2 – water 408, 430 UO2 – water system 431
y-component of the momentum equation 470 y-momentum equation 607, 628 Yuen and Theofanous 429
valve characteristics 358 valve performance 396 vdp-exergy 314 vector 637 vector (cross) products 653 vector between two points 637
Wall friction force 505 Wallis 421 wave 352 wave equation 477 weak pressure waves 409 weak waves 410 weighted average velocity 9 wet perimeters 73 Whitaker‘s spatial averaging theorem 50 x-component of the momentum equation 469 x-momentum equation 605, 625
z momentum equation 608 z-component of the momentum equation 472 Zeldovich 417 z-momentum equation 608, 630